Coverage boundary: reviewed FreeBench offers plus OpenRouter, Kilo, and OpenCode public catalogs. This is the complete published source list, not every offer on the internet. Catalogs refresh daily; failed checks preserve the last good snapshot and become stale after 48 hours. Changed provider terms await review.
Provider sources checked through 2026-11-03
Free access47 listings
Free access provider offers
Offer
Access route
Stated allowance
Before you connect
Source checked
Next step
cohere · Free inferenceCohere: North Mini Code (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: cohere/north-mini-code:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm cohere/north-mini-code:free is still listed and check the provider's current account and billing terms before use.
cohere/north-mini-code:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
cohere · Free inferenceCohere: North Mini Code (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: cohere/north-mini-code:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm cohere/north-mini-code:free is still listed and check the provider's current account and billing terms before use.
cohere/north-mini-code:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
dots-studio · Free inferenceDots Studio: Dots3-Note Preview (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: dots-studio/dots-3-note-preview:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
2026-12-31
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm dots-studio/dots-3-note-preview:free is still listed and check the provider's current account and billing terms before use.
dots-studio/dots-3-note-preview:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
dots-studio · Free inferenceDots Studio: Dots3-Note Preview (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: dots-studio/dots-3-note-preview:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
2026-12-31
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm dots-studio/dots-3-note-preview:free is still listed and check the provider's current account and billing terms before use.
dots-studio/dots-3-note-preview:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
Freebuff · Free agent or harness accessFreebuff free coding agentFull record and sources
Access details
Freebuff offers desktop and CLI coding-agent products funded by ads.
Harness fit
Freebuff Desktop and CLI workflows, pending integration feasibility.
Provider-stated limits
Freebuff advertises its coding agent at €0 forever, funded by ads. The public page reviewed does not state an exact daily request or token allowance; check the current account before use.
Eligibility
Account and regional eligibility details are not stated on the reviewed offer page; confirm availability in the current account.
Published expiry
Not stated; unknown.
What remains unknown
Model availability, regional eligibility, session limits, and data-use notices can vary. Freebuff says its agents can access files and environment data outside the project; its published privacy policy describes prompt and project-data review and ad personalization. A revised policy dated 2026-10-13 is scheduled to take effect 2026-11-12; review the current terms before sharing code. Benchmark integration is not qualified.
Practical next step
Confirm your account's visible allowance, regional eligibility, and model notices before using it for a workload.
google · Free inferenceGoogle: Gemma 4 26B A4B (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: google/gemma-4-26b-a4b-it:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm google/gemma-4-26b-a4b-it:free is still listed and check the provider's current account and billing terms before use.
google/gemma-4-26b-a4b-it:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
google · Free inferenceGoogle: Gemma 4 31B (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: google/gemma-4-31b-it:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm google/gemma-4-31b-it:free is still listed and check the provider's current account and billing terms before use.
google/gemma-4-31b-it:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
google · Free inferenceGoogle: Gemma 4 31B (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: google/gemma-4-31b-it:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm google/gemma-4-31b-it:free is still listed and check the provider's current account and billing terms before use.
google/gemma-4-31b-it:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
google · Free inferenceGoogle: Lyria 3 Clip PreviewFull record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: google/lyria-3-clip-preview.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm google/lyria-3-clip-preview is still listed and check the provider's current account and billing terms before use.
google/lyria-3-clip-preview · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
google · Free inferenceGoogle: Lyria 3 Pro PreviewFull record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: google/lyria-3-pro-preview.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm google/lyria-3-pro-preview is still listed and check the provider's current account and billing terms before use.
google/lyria-3-pro-preview · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
Kilo · Free inferenceKilo free model accessFull record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests.
Practical next step
Choose a specific :free model when a stable model identity matters; use Auto Free only when automatic model changes and its data terms are acceptable.
liquid · Free inferenceLiquidAI: LFM2.5-2.6B (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: liquid/lfm-2.5-2.6b:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm liquid/lfm-2.5-2.6b:free is still listed and check the provider's current account and billing terms before use.
liquid/lfm-2.5-2.6b:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
liquid · Free inferenceLiquidAI: LFM2.5-2.6B (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: liquid/lfm-2.5-2.6b:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm liquid/lfm-2.5-2.6b:free is still listed and check the provider's current account and billing terms before use.
liquid/lfm-2.5-2.6b:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
Nous Research · Free agent or harness accessNous Portal Free with HermesFull record and sources
Access details
Nous Portal's Free plan includes free models; Hermes is the available agent path.
Harness fit
Hermes agent connected through Nous Portal.
Provider-stated limits
The Free plan is listed at $0 with free models only, standard rate limits, and no monthly credits.
Eligibility
The reviewed plan page lists a Free plan; account and regional eligibility details are not specified here.
Published expiry
Not stated; unknown.
What remains unknown
The plan gives no universal request, token, or reset cap. Nous's privacy policy describes prompt and output processing for service development and training; Privacy Mode may limit Nous's own use prospectively, but third-party model-provider terms still apply.
Practical next step
Check the current Portal model picker and Hermes connection flow before selecting a model.
nvidia · Free inferenceNVIDIA: Nemotron 3 Nano Omni (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
nvidia · Free inferenceNVIDIA: Nemotron 3 Nano Omni (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
nvidia · Free inferenceNVIDIA: Nemotron 3 Ultra (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: nvidia/nemotron-3-ultra-550b-a55b:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3-ultra-550b-a55b:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3-ultra-550b-a55b:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
nvidia · Free inferenceNVIDIA: Nemotron 3 Ultra (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: nvidia/nemotron-3-ultra-550b-a55b:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3-ultra-550b-a55b:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3-ultra-550b-a55b:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
nvidia · Free inferenceNVIDIA: Nemotron 3.5 Content Safety (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: nvidia/nemotron-3.5-content-safety:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3.5-content-safety:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3.5-content-safety:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
nvidia · Free inferenceNVIDIA: Nemotron 3.5 Content Safety (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: nvidia/nemotron-3.5-content-safety:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3.5-content-safety:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3.5-content-safety:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
nvidia · Free inferenceNVIDIA: Nemotron 3.5 Lightning (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: nvidia/nemotron-3.5-lightning:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3.5-lightning:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3.5-lightning:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
nvidia · Free inferenceNVIDIA: Nemotron 3.5 Lightning (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: nvidia/nemotron-3.5-lightning:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3.5-lightning:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3.5-lightning:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
nvidia · Free inferenceNVIDIA: Nemotron 3.5 Super VL (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: nvidia/nemotron-3.5-super-vl:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3.5-super-vl:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3.5-super-vl:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
nvidia · Free inferenceNVIDIA: Nemotron 3.5 Super VL (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: nvidia/nemotron-3.5-super-vl:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nvidia/nemotron-3.5-super-vl:free is still listed and check the provider's current account and billing terms before use.
nvidia/nemotron-3.5-super-vl:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
OpenCode · Limited-time promotionOpenCode Zen listed-free modelsFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code.
Practical next step
Check the current model list, account billing controls and model-specific data policy; stop if the route is not $0 or a paid fallback is possible.
OpenCode v2 Console · select an exact model currently listed at $0.
Six prioritized models listed $0; OpenCode marks free models limited time.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: beam-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: beam-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm beam-free is still listed and check the provider's current account and billing terms before use.
beam-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: big-pickleFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: big-pickle.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm big-pickle is still listed and check the provider's current account and billing terms before use.
big-pickle · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: deepseek-v4.1-flash-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: deepseek-v4.1-flash-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm deepseek-v4.1-flash-free is still listed and check the provider's current account and billing terms before use.
deepseek-v4.1-flash-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: flint-40b-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: flint-40b-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm flint-40b-free is still listed and check the provider's current account and billing terms before use.
flint-40b-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: glm-5.3-flash-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: glm-5.3-flash-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm glm-5.3-flash-free is still listed and check the provider's current account and billing terms before use.
glm-5.3-flash-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: mimo-v2.6-flash-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: mimo-v2.6-flash-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm mimo-v2.6-flash-free is still listed and check the provider's current account and billing terms before use.
mimo-v2.6-flash-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: nemotron-3.5-lightning-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: nemotron-3.5-lightning-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nemotron-3.5-lightning-free is still listed and check the provider's current account and billing terms before use.
nemotron-3.5-lightning-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: nemotron-3.5-super-vl-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: nemotron-3.5-super-vl-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm nemotron-3.5-super-vl-free is still listed and check the provider's current account and billing terms before use.
nemotron-3.5-super-vl-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
OpenCode · Limited-time promotionOpenCode: pulse-alpha-freeFull record and sources
Access details
Select a currently listed free model through OpenCode v2 and verify the exact provider/model ID before use. Catalog model ID: pulse-alpha-free.
Harness fit
Native OpenCode v2 with the OpenCode Console provider; endpoint and protocol vary by model.
Provider-stated limits
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use. OpenCode currently lists this model in its free model catalog. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Eligibility is model-specific and may vary by account and region; verify the selected model in OpenCode Console.
Published expiry
Not stated; unknown.
What remains unknown
OpenCode currently describes all its free models as available for a limited time; this is a changing offer, not a permanent free tier. Model-specific data terms and regional eligibility differ. Beam Free has region restrictions and 30-day retention for safety; GLM-5.3-Flash Free and DeepSeek V4.1 Flash Free state zero retention and no training. The exact previously proposed `muse-spark-1.3-contributor-free` SKU is not listed; do not substitute the separately listed paid Muse model. Review the selected model's data terms and account billing/auto-recharge settings before sending code. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm pulse-alpha-free is still listed and check the provider's current account and billing terms before use.
pulse-alpha-free · OpenCode currently lists this model in its free model catalog.
As checked 2026-10-22, OpenCode lists the prioritized Zen routes `beam-free`, `big-pickle`, `deepseek-v4.1-flash-free`, `glm-5.3-flash-free`, `longcat-2.5-preview-free`, and `mimo-v2.6-flash-free` at zero input and output price. OpenCode describes its free models as available for a limited time. The Console docs publish no daily request or token quota; verify the exact ID and account eligibility before use.
Free access can change; verify the exact model, region, and billing settings.
openrouter · Free inferenceFree Models RouterFull record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: openrouter/free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm openrouter/free is still listed and check the provider's current account and billing terms before use.
openrouter/free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
openrouter · Free inferenceOpenRouter Free Models RouterFull record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: openrouter/free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm openrouter/free is still listed and check the provider's current account and billing terms before use.
openrouter/free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
OpenRouter · Free inferenceOpenRouter free model variantsFull record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure.
Practical next step
Read your key's free-model daily counter before a long run, and treat a 429 as an exhausted allowance rather than a model result.
recursive · Free inferenceRecursive: Flint (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: recursive/flint-v1.0:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm recursive/flint-v1.0:free is still listed and check the provider's current account and billing terms before use.
recursive/flint-v1.0:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
recursive · Free inferenceRecursive: Flint (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: recursive/flint-v1.0:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm recursive/flint-v1.0:free is still listed and check the provider's current account and billing terms before use.
recursive/flint-v1.0:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
reflection · Free inferenceReflection: Beam 1.0 (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: reflection/beam-free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm reflection/beam-free is still listed and check the provider's current account and billing terms before use.
reflection/beam-free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
stealth · Free inferenceStealth: Glyph Cluster (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: stealth/glyph-cluster.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm stealth/glyph-cluster is still listed and check the provider's current account and billing terms before use.
stealth/glyph-cluster · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
thinkingmachines · Free inferenceThinking Machines: Inkling (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: thinkingmachines/inkling:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
2026-11-04
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm thinkingmachines/inkling:free is still listed and check the provider's current account and billing terms before use.
thinkingmachines/inkling:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
thinkingmachines · Free inferenceThinking Machines: Inkling (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: thinkingmachines/inkling:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
2026-11-04
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm thinkingmachines/inkling:free is still listed and check the provider's current account and billing terms before use.
thinkingmachines/inkling:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
thinkingmachines · Free inferenceThinking Machines: Inkling Small (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: thinkingmachines/inkling-small:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
2026-11-04
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm thinkingmachines/inkling-small:free is still listed and check the provider's current account and billing terms before use.
thinkingmachines/inkling-small:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
thinkingmachines · Free inferenceThinking Machines: Inkling Small (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: thinkingmachines/inkling-small:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
2026-11-04
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm thinkingmachines/inkling-small:free is still listed and check the provider's current account and billing terms before use.
thinkingmachines/inkling-small:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
xingchen-agi · Free inferenceXingChen: Xing4.0 29B A4B (free)Full record and sources
Access details
Call any model whose ID ends in :free through the OpenRouter API or an OpenAI-compatible gateway client. Catalog model ID: xingchen-agi/xing4.0-29b-a4b:free.
Harness fit
Any OpenAI-compatible client, including agent harnesses that call OpenRouter directly.
Provider-stated limits
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining. OpenRouter model ID listed at zero prompt and completion price. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Free model variants are available through OpenRouter accounts; the daily cap depends on all-time credits purchased, as stated in the published limits page.
Published expiry
Not stated; unknown.
What remains unknown
Eligible :free model IDs, per-model availability and the daily ceiling can change, and the per-minute limit is not reported by the key endpoint. Rate-limit headers appear only on a 429 response, so the account counter has to be read before a long run to tell an exhausted allowance from a model failure. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm xingchen-agi/xing4.0-29b-a4b:free is still listed and check the provider's current account and billing terms before use.
xingchen-agi/xing4.0-29b-a4b:free · OpenRouter model ID listed at zero prompt and completion price.
Free variants are capped at 20 requests per minute and 50 requests per day for accounts that have purchased fewer than 10 credits, rising to 1,000 requests per day at 10 or more all-time credits purchased. Capacity is governed globally, so additional accounts or API keys do not raise these limits. GET https://openrouter.ai/api/v1/key reports free_model_daily_requests with used, limit and remaining.
Check the account counter; the daily allowance is shared across free models.
xingchen-agi · Free inferenceXingChen: Xing4.0 29B A4B (free)Full record and sources
Access details
Kilo provides two ways to use free models: choose an exact model tagged :free in the model picker/Gateway, or select Auto Free (`kilo-auto/free`) to route requests automatically among available free models. These are two routing modes under one Kilo access offer. Catalog model ID: xingchen-agi/xing4.0-29b-a4b:free.
Harness fit
Kilo Code and CLI; exact-model Gateway clients can select a fixed model, while Auto Free uses Kilo-managed model selection.
Provider-stated limits
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply. Kilo Gateway catalog lists this model as free. Published request, token, and reset limits that are not in this model record remain unknown.
Eligibility
Kilo documents the free Gateway allowance for authenticated and anonymous access; upstream model provider eligibility and regional availability may also apply.
Published expiry
Not stated; unknown.
What remains unknown
Free model IDs and availability change. Auto Free does not promise a fixed underlying model and may route to providers that log prompts and outputs or use them to improve their services, including NVIDIA trial endpoints. Do not submit personal or confidential data through Auto Free. Kilo's autocomplete and background tasks have separate settings and may use credits; this listing covers free model access for agentic requests. This individual model record does not state separate request, token, or reset quotas.
Practical next step
Confirm xingchen-agi/xing4.0-29b-a4b:free is still listed and check the provider's current account and billing terms before use.
xingchen-agi/xing4.0-29b-a4b:free · Kilo Gateway catalog lists this model as free.
Kilo documents zero cost for Gateway models tagged `:free` and a 200 requests-per-hour-per-IP limit for free model requests, for both authenticated and anonymous access. Auto Free routes to currently available free models; upstream limits may also apply.
Auto Free changes models automatically and may send prompts to providers that log them.
OpenCode · Paid plan with free modelsOpenCode Go subscriptionFull record and sources
Access details
Subscribe to Go or Go Plus in OpenCode Console, then connect the provider from the TUI.
Harness fit
Native OpenCode v2 with the OpenCode Go provider.
Provider-stated limits
Go is $10/month and Go Plus is $40/month. Usage limits are monthly dollar amounts per model, with separate 5-hour, weekly, and monthly windows. The current Go docs list `beam-free` and `longcat-2.5-preview-free` as free for a limited time with unlimited usage; this remains a paid subscription, not free access.
Eligibility
Eligibility details are not stated; unknown.
Published expiry
Not stated; unknown.
What remains unknown
Subscription usage is not a free tier. Model availability and data terms vary; some model entries disclose training use. Review the selected model and account fallback controls.
Practical next step
Compare the current subscription terms before using Go as a paid upgrade path.