Free AI models, by workload

Pick the job you're building. Each list filters and orders free-listed catalog records using provider-declared context windows, tool support and price. The rule is stated on each page; it describes catalog facts, not task performance.

Ranked from declared catalog facts — not from measured benchmark runs. For measured outcomes, check the exact model × harness reports in the evaluation register.

• Coding agent

  Ship a repo-wide change: edit files, run commands, fix its own breakage.

  16 free-listed models · cost weight 0.15

• Tool pipeline

  Multi-step automation that must call the right tools, in the right order, every time.

  16 free-listed models · cost weight 0.15

• Long-context research

  Read a large corpus, keep the thread, and summarise without losing the details.

  19 free-listed models · cost weight 0.20

• Customer-facing automation

  Runs without a human watching: confirm before destructive work and never cave to bad instructions.

  19 free-listed models · cost weight 0.10

• High-volume batch

  Thousands of short tasks where cost per success, not the peak score, decides.

  19 free-listed models · cost weight 0.60

• Zero-budget (free only)

  Free model variants only — the FreeBench niche, weighted for general agent work.

  19 free-listed models · cost weight 0.40

• Orchestrator / planner

  A lead agent that plans, delegates to sub-agents, and merges their results.

  19 free-listed models · cost weight 0.15