// the find
xtekky/gpt4free
The official gpt4free repository | various collection of powerful language models | opus 4.6 gpt 5.3 kimi 2.5 deepseek v3.2 gemini 3
gpt4free is a Python library that routes LLM requests through scraped or reverse-engineered provider endpoints, letting you call GPT-4, Claude, Gemini, and others without paying for API access. It wraps everything in an OpenAI-compatible API surface so existing code mostly just works. The target audience is hobbyists and developers who want to experiment with LLMs without billing.
The OpenAI-compatible Interference API is genuinely useful — you can point existing tooling at localhost:1337 and most things work without code changes. Provider breadth is wide: 50+ backends including local inference via Ollama, image generation via Pollinations, and browser-automated providers using HAR files. The Docker setup is well-thought-out, with a VNC-accessible container for providers that need interactive login. The MCP server integration is a nice recent addition that makes it usable as a tool source for Claude Desktop without much configuration.
The entire premise is legally precarious — most providers explicitly forbid automated access in their ToS, and the LEGAL_NOTICE.md in the repo is essentially a disclaimer that the maintainers know this. Any provider can break without warning when they rotate tokens or change their API shape, and there's no SLA on fixes. For production use the reliability story is bad: you're dependent on scraping fragile third-party endpoints, and the authentication flow for browser-based providers (HAR files, cookies) rots fast. The codebase has grown organically to 50+ provider files with no consistent interface discipline — reading one provider tells you almost nothing about how the next one works.