Skip to content

Free LLM APIs

“Free” means something different at every provider

The limits, whether you need a card, who may read or train on your prompts, and where you're allowed to use it. Read on each provider's own pages, with links and dates.

Read on each provider's own documentation and terms; oldest check 2026-10-22, and each page shows its date and sources. Providers change these often: follow the links before relying on a number.

Five things to check before you build on a free tier

Is it a free tier or a trial?

Credits that expire after 30 days are a trial. A free tier renews every day.

Does it need a card?

With a card on file, some accounts move to paid usage without you noticing. Check the billing page, not the pricing page.

Who reads your prompts?

Some free tiers pay for themselves with your data: used to improve products, read by reviewers, or used by the provider serving the model to train.

Where may you use it?

Some free tiers can't be used in apps for people in certain regions: the Gemini API free tier is not allowed for users in the EEA, Switzerland or the UK.

What's the real limit?

Requests per day, but also tokens per minute: a long prompt can fail on the first request of the day.

State of free LLM APIs

How many free models there are today, how good they are, and how many go to providers that may train on your prompts. Updated daily.

The open ranking on GitHub

Every free model ranked daily from published benchmarks and LM Arena, with the scoring code and the data as JSON. MIT.

We maintain that ranking and run ZeroLimitAI, an OpenAI-compatible endpoint that sends each request to the best free model available, and can use your own Groq or OpenRouter key.

Stop paying per token.

One endpoint. $0 inference.

Free 7-day API key in a minute — no card. Then $49 a year or $99 once.