Free LLM API limits

These notes copy what each provider’s own docs said on 22 October 2026. A console can show a different quota for your account. Where a doc does not publish a number, the provider page says so instead of filling one in. GitHub Models is included because the docs now say the product is retired.

Provider	What “free” means in the docs
Gemini API	No public free-tier RPM table. Free spend cap is N/A. Unpaid quota has its own data terms.
Groq	Free Plan Limits payload, plus a sentence on the same page that calls the visible table Developer-plan base limits.
Cerebras	$5 for 30 days after a verified card. No renewing free tier. Two official pages disagree on gpt-oss-120b RPM.
Mistral	Free mode is the default account. RPM and TPM are on a logged-in Limits page.
Cohere	Trial keys: 1,000 calls a month, and 20 requests per minute on the listed chat models.
OpenRouter	:free variants: 20 requests per minute, and 50 or 1,000 requests per day.
GitHub Models	Fully retired as of July 30, 2026. No current rate table.
Workers AI	10,000 Neurons per day, then $0.011 per 1,000 Neurons on Workers Paid.

Free LLM API limits

Directory entries

Sources