Providers that let you call a large language model without paying. Some are free forever, some hand out signup credits, and some are limited-time offers — rated 0-5 on quota, rate limits and stability to help you pick quickly.
Free tiers usually come with strict limits. They are best suited to development, prototyping and testing — we do not recommend relying on them in production.
Quota, rate limit and stability ratings are subjective assessments by the site editors based on hands-on testing. They are indicative only — always check the provider's own documentation for the actual terms.
8 active entries
Google AI Studio provides a free tier for Gemini models on each account, with request quotas reset daily. Different models have different limits; please refer to the actual limits for specifics.
Some of Ollama's cloud models offer a certain amount of free quota, with relatively relaxed QPS limits, and the quota resets every two weeks. Registration may require overseas phone number verification.
A unified entry point aggregating hundreds of models. Only models marked with the :free suffix can be used for free. Providers often offer limited-time free access to new models on OpenRouter, but free models may be slower and less stable.
Cerebras offers some free models to developers, with a $5 balance gift upon registration. Note: Cerebras does not support registration and use by users in China.
Each Cloudflare account gets 10k neurons free per day, can call some open-source models, a bit complex to get started, average speed.
AMD Radeon Cloud is a free shared LLM API service provided by AMD for developers, compatible with the OpenAI interface. Due to the large number of users, speed and stability are average.
NVIDIA's official free large model service, requires mobile phone verification during registration, has relatively strict quotas, and moderate stability.
Groq is an American AI chip and cloud service company (not Musk's Grok, mind you), known for its LPU chips designed to accelerate large-model inference and its high-speed inference cloud platform. It offers free quotas for some open-source models; the speed is very fast, but the quota is relatively small, the models are relatively old, and it resets once a day.
We collect publicly documented free tiers, signup credits and limited-time offers from LLM providers, then summarise what you get and how usable it is. It is manually curated and updated as offers change.
Unofficial directory, not affiliated with any provider listed. Free offers change frequently and may be region-restricted. Always verify the current terms on the provider's own website — the link on each card is the authoritative source.
