The free LLM API directory that could save your wallet
A living spreadsheet of which AI providers actually let you call their models for free, with rate limits and gotchas spelled out.

What it does
This repo is a curated, auto-updated directory of services offering free API access to large language models. It splits providers into two camps: genuinely free tiers (OpenRouter, Google AI Studio, Groq, Cloudflare Workers AI, and others) and those handing out trial credits (Fireworks, Baseten, Nebius, etc.). Each entry lists the specific models available and their rate limits, so you can compare quotas without clicking through a dozen pricing pages.
The interesting bit
The README is generated by a Python script (src/pull_available_models.py), meaning the list is meant to stay current as providers add or remove free models. The author also explicitly excludes reverse-engineered chatbot wrappers — a small act of curation that keeps the list legally and practically useful.
Key highlights
- Covers 13+ free providers and 14+ trial-credit providers
- Rate limits are granular: per-minute, per-day, per-month, and per-model where applicable
- Includes big names (Google, NVIDIA, Mistral, GitHub Models) and smaller gateways (OpenCode Zen, Vercel AI Gateway)
- Warns when free tiers feed your data into training (Google AI Studio outside the EU, Mistral’s Experiment plan)
- Some limits are quirky: Cloudflare uses “neurons/day,” GitHub Models ties access to your Copilot subscription tier
Caveats
- The auto-generation claim is in the header comment, but the repo doesn’t expose how often it actually runs
- Several providers require phone verification (NVIDIA NIM, Mistral) or have vague model lists (“various open models”)
- Rate limits change; the README captures a snapshot, not a guarantee
Verdict
Worth bookmarking if you’re prototyping, hacking on side projects, or just refuse to pay $20 to test an API. Skip it if you need production SLAs or predictable costs — free tiers are free for a reason.