Skip to main content
Back

FreeLLMAPI Stacks 16 Free AI Tiers Into 1.7B Monthly Tokens

FreeLLMAPI is a self-hosted TypeScript proxy that stacks free-tier keys from 16 AI providers behind one OpenAI-compatible endpoint, adding up to ~1.7B tokens per month with automatic failover.

Repo, Full Access
  • What it is: FreeLLMAPI is a self-hosted, MIT-licensed TypeScript proxy that stacks free-tier keys from 16 AI providers behind one OpenAI-compatible endpoint.
  • Token volume: Combined free tiers add up to ~1.7 billion tokens per month across 100+ models, with Mistral alone contributing roughly 1 billion.
  • Providers supported: Google, Groq, Cerebras, SambaNova, NVIDIA, Mistral, OpenRouter, GitHub Models, Hugging Face, Cohere, Cloudflare, Zhipu, Moonshot, MiniMax, and more.
  • Key features: Automatic failover on 429s/5xxs, AES-256-GCM encrypted key storage, sticky sessions, per-key rate tracking, and a React admin dashboard.
  • Limitations: No tool calling, no vision/multimodal, no embeddings yet; effective model quality degrades throughout the day as top-tier daily caps are exhausted.
  • Scope: Personal experimentation only -- not designed for production or multi-tenant use; Cohere and NVIDIA NIM have ToS restrictions to be aware of.

FreeLLMAPI is a self-hosted, MIT-licensed TypeScript proxy built by tashfeenahmed that solves a surprisingly common frustration: every major AI lab now offers a free API tier, but each one is too small to be useful on its own. The project's answer is to stack them. 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 1.7 billion tokens per month of working inference capacity, across 100+ models from small-and-fast to reasonably capable.

The repo has quietly accumulated nearly 10,000 stars, and the pitch is simple: an OpenAI-compatible proxy that stacks the free tiers of 16 LLM providers behind one /v1 endpoint, with smart routing, automatic failover, and encrypted keys. You change one line of code -- base_url -- and your existing OpenAI SDK calls start routing across a fleet of free providers transparently.

Fourteen providers, one endpoint

The proxy aggregates free tiers from Google, Groq, Cerebras, NVIDIA, Mistral, OpenRouter, GitHub Models, Cohere, Cloudflare, HuggingFace, Zhipu, Moonshot, MiniMax, and SambaNova -- plus any custom OpenAI-compatible endpoint like llama.cpp, LM Studio, vLLM, or a local Ollama. The provider list has grown since the initial release as contributors add adapters.

FreeLLMAPI fallback chain configuration showing model priority, token budgets, and performance metrics

The free-tier token math is real, though uneven. Mistral alone contributes roughly 1 billion tokens; everything else is a bonus. Providers like Groq and Cerebras add speed rather than raw volume -- Cerebras and Groq are extremely fast; others are not. You get whichever one is available.

How the routing actually works

FreeLLMAPI is an API proxy and load balancer designed to aggregate the free-tier offerings of multiple LLM providers into a single, unified, OpenAI-compatible endpoint. It provides a transparent routing layer that handles model selection, automatic fallbacks, and per-key usage tracking.