Free LLM API providers
Every monitored platform on one page: who has the most free models, who needs no card, and which base URL to set in your client.
Updated:
| Provider ↕ | Models ↕ | Best free model | Card | Base URL | |
|---|---|---|---|---|---|
 OpenRouter Models with the :free suffix allow 50 requests per day, or 1,000 per day after a one-time $10 top-up (20 requests per minute). |
21 | Qwen3.8 27B | No | https://openrouter.ai/api/v1 |
Key |
 Groq Free plan with per-model limits on requests and tokens per minute and per day; very high generation speed on LPU hardware. |
4 | Qwen3.8 27B | No | https://api.groq.com/openai/v1 |
Key |
 Google AI Studio The Gemini API free tier covers Flash and Flash-Lite models with rate limits; free-tier content may be used to improve Google products. |
4 | Gemini 2.0 Flash | No | https://generativelanguage.googleapis.com/v1beta/openai/ |
Key |
 Mistral AI The Free plan includes monthly API credits for Mistral models, with low default rate limits. |
4 | Mistral Nemo | — | https://api.mistral.ai/v1 |
Key |
 Together AI Only selected models are priced at zero; there is no general free trial. |
98 | GLM 5.3 | — | https://api.together.xyz/v1 |
Key |
 Cohere Trial keys allow 1,000 API calls per month for evaluation, not for production. |
4 | Command R Plus | No | https://api.cohere.ai/compatibility/v1 |
Key |
 Cerebras Free trial with daily token limits; a verified payment method is required to activate the API. |
2 | Qwen 3.8 27B | Yes | https://api.cerebras.ai/v1 |
Key |
 Cloudflare Workers AI 10,000 Neurons per day are free on both Workers Free and Paid plans. |
3 | Llama 3.1 8B | — | https://api.cloudflare.com/client/v4/accounts/ACCOUNT_ID/ai/v1 |
Key |
 Hugging Face Free accounts get a small monthly credit for Inference Providers; PRO accounts get more. |
3 | Llama 3.1 8B | No | https://router.huggingface.co/v1 |
Key |
 OrcaRouter BYOK router without token markup; its own Fusion models and selected free routes cost nothing. |
13 | GLM 5.3 Flash | — | https://api.orcarouter.ai/v1 |
Key |
 NVIDIA Build Free serverless inference for development and prototyping with a developer account; not intended for production traffic. |
62 | GLM 5.3 | No | https://integrate.api.nvidia.com/v1 |
Key |
 TokenRouter Free models and promotional access may be time-limited. |
1 | Nemotron 3 Nano Omni | — | https://api.tokenrouter.com/v1 |
Key |
 Vercel AI Gateway A payment card must be on file even for zero-price models. |
3 | Laguna S 2.1 | Yes | https://ai-gateway.vercel.sh/v1 |
Key |
 Nous Portal The $0 Portal plan includes :free models with standard rate limits. |
7 | Solar Pro 4 | No | https://inference-api.nousresearch.com/v1 |
Key |
 UnoRouter Large catalog of :free routes without a card; shared capacity and per-model limits. |
62 | Mimo V2.6 Pro | No | https://api.unorouter.com/v1 |
Key |
FAQ
Which free LLM API provider is best?
It depends on the job. OpenRouter and UnoRouter give the most models behind one key, Groq and Cerebras are the fastest, Google AI Studio has generous Gemini Flash limits, and NVIDIA Build hosts fresh open models for development.
Can I combine several free providers?
Yes. They are all OpenAI-compatible, so clients such as LiteLLM, OpenCode, or Open WebUI let you switch keys and stack free quotas.