Methodology

What's included

LLM API providers that let you call models without paying: free tiers with rate limits, individual free models (like OpenRouter's :free variants) and free monthly credit or quotas (66 models and offers from 13 providers). One-off sign-up credits that expire, chat apps without an API, and models you have to host yourself are left out. Services that stopped their free offer are listed under "Gone or not free anymore" on the home page.

Where the numbers come from

• Only the provider's own pages and docs. Each provider page lists the exact URLs and the date we checked them; all are on the sources page.
• Numbers are copied as published: requests per minute and per day, tokens per minute and per day. Anything else (per hour, per month, Neurons, audio minutes) goes into "other limits".
• If a provider doesn't publish a number publicly (Google and Mistral show limits only in their consoles), the table shows n/p and the API has null. We never estimate.
• "Card needed: no" is shown only when the provider says so. Otherwise it reads "not stated".

What it can't tell you

• Free tiers change without notice. Data was last checked 2026-10-25; confirm on the provider's page before you build on it.
• Published limits are maximums. Free requests may be slower, queued or dropped under load, and the limits may differ for your account or region.
• Terms of use apply: several free tiers are for evaluation only, not production, and some providers use free-tier prompts to improve their models.

Paid prices

For per-token prices once you outgrow the free tier, see prices.branolabs.com.

Data

The full dataset is at /api/free-tiers.json (CC BY 4.0) and summarised for AI assistants in /llms.txt. Found something wrong or outdated? Use the bug report link in the footer.