7.4 billion tokens per month. 34 free LLM providers. 635 free model endpoints. All behind one /v1 endpoint, plus any custom OpenAI-compatible endpoint. Smart routing, automatic failover, encrypted keys. Personal experimentation only.
TypeScript
Merge pull request #604 from tashfeenahmed/feat/modelscope-provider
Tashfeen145e40fe09f6
diff --cc README.mdindex 18daa13c,c4f9208f..911ee995--- a/README.md+++ b/README.md@@@ -2,7 -2,7 +2,7 @@@# FreeLLMAPI- **4 billion tokens per month. 28 free LLM providers. 339 free model endpoints. One OpenAI-compatible endpoint.**-**One OpenAI-compatible endpoint. 29 free LLM providers. 339 free model endpoints. ~4 billion tokens per month.**++**4 billion tokens per month. 29 free LLM providers. 358 free model endpoints. One OpenAI-compatible endpoint.**Aggregate free tiers from dozens of providers, plus custom OpenAI-compatible chat, embedding, image, and audio endpoints, behind a single `/v1` API. Keys are stored encrypted. A router picks the best available model for each request, falls over to the next provider when one is rate-limited, and tracks per-key usage so you stay under every free-tier cap.@@@ -13,10 -13,7 +13,10 @@@[](https://github.com/tashfeenahmed/freellmapi/pkgs/container/freellmapi)[](https://deepwiki.com/tashfeenahmed/freellmapi)--**[freellmapi.co](https://freellmapi.co/?utm_source=github&utm_medium=readme&utm_campaign=repository&utm_content=readme_top)** · browse the full catalog: 235 model families, 339 free endpoints++**[freellmapi.co](https://freellmapi.co/?utm_source=github&utm_medium=readme&utm_campaign=repository&utm_content=readme_top)** · browse the full catalog: 251 model families, 358 free endpoints+++Your router updates its own model catalog from a signed feed: new free models, quota changes, and compatibility fixes land without a `git pull`.**[Go live at freellmapi.co](https://freellmapi.co/?utm_source=github&utm_medium=readme&utm_campaign=premium&utm_content=readme_top#pricing)** ($19/yr, cancel anytime).@@@ -40,48 -39,49 +40,51 @@@- [Using the API](#using-the-api)- [Screenshots](#screenshots)- [How it works](#how-it-works)-- [Context Handoff](#context-handoff)- [Limitations](#limitations)-- [Documentation index](docs/README.md)-- [Contributor guide](CONTRIBUTING.md)- [Contributing](#contributing)-- [Terms of Service review](#terms-of-service-review)- [Disclaimer](#disclaimer)+**Guides:** [Install & deploy](docs/install.md) · [API reference](docs/api.md) · [Clients & coding agents](docs/clients.md) · [Prompt compression](docs/compression.md) · [Architecture & internals](docs/architecture.md) · [Documentation index](docs/README.md) · [Contributor guide](CONTRIBUTING.md)+## Why this exists- Every serious AI lab now offers a free tier, a few million tokens a month, a few thousand requests a day. On its own each tier is a toy. Stacked together, they add up to roughly **4 billion tokens per month** of working inference capacity, across **235 model families / 339 provider endpoints** from small-and-fast to reasonably capable.-Every serious AI lab now offers a free tier — a few million tokens a month, a few thousand requests a day. On its own each tier is a toy. Stacked together, they add up to roughly **4 billion tokens per month** of working inference capacity, across **235 model families / 339 provider endpoints** from small-and-fast to reasonably capable.++Every serious AI lab now offers a free tier, a few million tokens a month, a few thousand requests a day. On its own each tier is a toy. Stacked together, they add up to roughly **4 billion tokens per month** of working inference capacity, across **251 model families / 358 provider endpoints** from small-and-fast to reasonably capable.- The problem is that stacking them by hand is painful: twenty-eight different SDKs, twenty-eight different rate limits, twenty-eight places a request can fail. FreeLLMAPI collapses that into one OpenAI-compatible endpoint. Point any OpenAI client library at your local server, and it routes transparently across whichever providers you've added keys for.+ The problem is that stacking them by hand is painful: twenty-nine different SDKs, twenty-nine different rate limits, twenty-nine places a request can fail. FreeLLMAPI collapses that into one OpenAI-compatible endpoint. Point any OpenAI client library at your local server, and it routes transparently across whichever providers you've added keys for.And the free-tier landscape shifts weekly: providers launch models, retire them, and change quotas without notice. FreeLLMAPI tracks all of that for you. The router pulls a signed model catalog from [freellmapi.co](https://freellmapi.co) on its own, so your install keeps up without a `git pull`. See [Premium (live catalog)](#premium-live-catalog) for how fast it keeps up.++## Supported providers+<div align="center"><table><tr>-<td align="center" width="180"><a href="https://ai.google.dev"><b>Google</b><br/>Gemini 2.5 Flash · 3.x previews</a></td>-<td align="center" width="180"><a href="https://groq.com"><b>Groq</b><br/>Llama 3.3, Llama 4, GPT-OSS, Qwen3</a></td>-<td align="center" width="180"><a href="https://cerebras.ai"><b>Cerebras</b><br/>Qwen3 235B</a></td>-<td align="center" width="180"><a href="https://opencode.ai/zen"><b>OpenCode Zen</b><br/>DeepSeek V4 Flash · Nemotron (promo)</a></td>+<td align="center" width="150"><img src="repo-assets/providers/google.png" width="44" alt="Google"><br/><b>Google</b></td>+<td align="center" width="150"><picture><source media="(prefers-color-scheme: dark)" srcset="repo-assets/providers/groq-dark.png"><img src="repo-assets/providers/groq.png" width="44" alt="Groq"></picture><br/><b>Groq</b></td>+<td align="center" width="150"><img src="repo-assets/providers/cerebras.png" width="44" alt="Cerebras"><br/><b>Cerebras</b></td>+<td align="center" width="150"><picture><source media="(prefers-color-scheme: dark)" srcset="repo-assets/providers/opencode-dark.png"><img src="repo-assets/providers/opencode.png" width="44" alt="OpenCode Zen"></picture><br/><b>OpenCode Zen</b></td></tr><tr>-<td align="center"><a href="https://mistral.ai"><b>Mistral</b><br/>Large 3 · Medium 3.5 · Codestral · Devstral</a></td>-<td align="center"><a href="https://openrouter.ai"><b>OpenRouter</b><br/>21 free-tier models</a></td>-<td align="center"><a href="https://github.com/marketplace/models"><b>GitHub Models</b><br/>GPT-4.1 · GPT-4o</a></td>-<td align="center"><a href="https://developers.cloudflare.com/workers-ai"><b>Cloudflare</b><br/>Kimi K2 · GLM-4.7 · GPT-OSS · Granite 4</a></td>+<td align="center"><img src="repo-assets/providers/mistral.png" width="44" alt="Mistral"><br/><b>Mistral</b></td>+<td align="center"><img src="repo-assets/providers/openrouter.png" width="44" alt="OpenRouter"><br/><b>OpenRouter</b></td>+<td align="center"><img src="repo-assets/providers/cloudflare.png" width="44" alt="Cloudflare"><br/><b>Cloudflare</b></td>+<td align="center"><img src="repo-assets/providers/cohere.png" width="44" alt="Cohere"><br/><b>Cohere</b></td></tr><tr>-<td align="center"><a href="https://cohere.com"><b>Cohere</b><br/>Command R+ · Command-A (trial)</a></td>-<td align="center"><a href="https://docs.z.ai"><b>Z.ai (Zhipu)</b><br/>GLM-4.5 · GLM-4.7 Flash</a></td>-<td align="center"><a href="https://build.nvidia.com"><b>NVIDIA</b><br/>NIM · 40 RPM free (eval-only ToS)</a></td>-<td align="center"><a href="https://huggingface.co/docs/inference-providers"><b>HuggingFace</b><br/>Router → DeepSeek V4 · Kimi K2.6 · Qwen3</a></td>+<td align="center"><img src="repo-assets/providers/zhipu.png" width="44" alt="Z.ai (Zhipu)"><br/><b>Z.ai (Zhipu)</b></td>+<td align="center"><img src="repo-assets/providers/nvidia.png" width="44" alt="NVIDIA"><br/><b>NVIDIA</b></td>+<td align="center"><img src="repo-assets/providers/huggingface.png" width="44" alt="HuggingFace"><br/><b>HuggingFace</b></td></tr>+ <tr>+ <td align="center"><a href="https://modelscope.cn"><b>ModelScope</b><br/>Qwen3 · DeepSeek V4 · GLM-5 (needs Aliyun cn binding)</a></td>+ </tr></table>-<p align="center"><strong>…and many more.</strong></p>+<i>… and 17 more free providers</i>++</div>Plus a **custom** provider — point chat, embedding, image, or audio models at any OpenAI-compatible endpoint (llama.cpp, LM Studio, vLLM, a local Ollama, or a remote gateway) from the Keys page.@@@ -237,7 -429,7 +240,7 @@@ quota changes, and provider quirk fixeenable/disable choices and custom providers are never touched, and everydownload is verified against a pinned Ed25519 key before it is applied.--The catalog currently tracks **28 providers**, **235 model families**, **339++The catalog currently tracks **29 providers**, **251 model families**, **358provider/model endpoints**, and roughly **4 billion tokens per month** of listedfree-tier capacity. Browse the full set at**[freellmapi.co/models](https://freellmapi.co/models.html)**.