Free LLM API limits compared
6 providers with a recurring free tier, no card required. Free tiers change often: Groq retired its free Llama models in August 2026, and Cerebras now offers only a one-time trial credit. Need many models behind one key? Use OpenRouter :free and fall back across providers when you hit a limit.
Hosted APIs with a free tier or $0 routes. A dash means the provider does not publish that number — see the notes column. Start here for RPM and daily caps, then open a provider page or a model free API guide. Credits and timed deals are on deals.
| Provider | RPM | Daily requests | Daily tokens | Notes | Checked |
|---|---|---|---|---|---|
| Cloudflare Workers AI | — | — | — | 10,000 Neurons/day shared across all models — not a request count. Bigger models burn more Neurons per call. | 2026-10-10 |
| Google Gemini | — | — | — | Google no longer publishes fixed free-tier RPM/RPD. Your project's limits are shown in AI Studio → Usage & limits. | 2026-10-10 |
| Groq | 30 | 1000 | 200000 | 2026-10-10 | |
| Mistral AI | — | — | — | Free mode limits are per organization (Admin → Limits) and not published. Free keys can call Nemo, Ministral and Codestral, not mistral-small or Large. | 2026-10-10 |
| NVIDIA NIM | 40 | — | — | Per-model trial limit; NVIDIA does not publish it. Your cap is shown on build.nvidia.com. Prototyping only, not production. | 2026-10-10 |
| OpenRouter | 20 | 50 | — | 1,000 :free requests/day after buying $10 in credits. | 2026-10-10 |
Limits checked against official docs on 2026-10-10. Rates change without notice. Not affiliated with providers unless noted. Verify before production use.