AI Rate CardFree AI API: models you can call for free today, and the catch
Updated Oct 19, 2026
In the 19 October 2026 collection 155 offers of text models cost nothing per token. Most are "free" variants that routers and hosts publish next to the paid one (OpenRouter's :free, OpenCode's "Free", AIHubMix's "(free)"), a few are free endpoints of the model maker itself (NVIDIA). The table below is rebuilt from today's catalog on every page view: one row per model, the longest context first.
Free text models right now
Live data: this table is built from today's prices.
What "free" means
- Rate limits. Free variants are throttled: a few requests per minute and a daily cap are typical, and they are the first to be slowed at peak hours. Good for trying a model or a hobby project, not for production.
- Your data. Several free variants say requests may be logged or used to improve the model. Do not send anything private through them.
- They come and go. Free variants appear at a model's launch and disappear without notice. Every row links to a card with the date it was last checked.
- Same model, different conditions. A free GLM or MiniMax "coding" variant can be limited to coding tools or a shorter context than the paid one - read the provider's note.
When free stops being enough
The cheapest paid text models cost a few cents per million tokens - for most projects less than a dollar a month. The live list is in the cheapest LLM API, and the cost calculator turns your monthly volume into a price at every provider.
All prices in US dollars, as the providers publish them; a free offer is marked "Free" in the catalog and on its card.