gpt4free

A community run Python project that routes requests across many language model providers, with a local GUI, an OpenAI compatible API, a JavaScript client, and Docker images. Licensed under GPLv3.

The repository started as a single place to reach different language models. Its description on GitHub calls it an official collection of powerful models, and the README frames gpt4free as a community driven effort that aggregates accessible providers and interfaces. The point is to make modern LLMs and media generation easier to work with through one tool instead of juggling separate accounts and endpoints. The project is Python under the hood, released under GPLv3, started by @xtekky and maintained day to day by @hlohaus.

The README lists the components: a Python client library with an async client, an optional local web GUI, a FastAPI based OpenAI compatible API called the Interference API, an official browser JavaScript client, Docker images in full and slim variants, multi provider adapters, and tooling for image, audio, and video generation. Requirements include Python 3.10 or later and Google Chrome or Chromium for providers that rely on browser automation.

Docker is the recommended route, with persistent directories and port 8080 serving the GUI and API. The slim image supports both x64 and arm64. A Windows guide walks through downloading the g4f.exe release artifact, unzipping it, and running g4f.exe, then opening the chat UI at localhost:8080. Python installs are covered as well, via pip, from source, or with partial installs for a smaller footprint.

The Interference API is an OpenAI compatible endpoint with a Swagger UI and documented defaults. A Model Context Protocol server exposes tools for web search through DuckDuckGo, web scraping, and image generation from text prompts. Providers integrate a broad mix: OpenAI compatible endpoints, PerplexityLabs, Gemini, MetaAI, Pollinations, and local inference backends. Depending on the provider, the adapter may need API keys, browser cookies, HAR files, or Chrome and Chromium tooling.

The security section is direct: do not store sensitive credentials, secure the server with HTTPS, authentication, and firewall rules, and limit access to cookie and HAR storage. Takedown requests go to takedown@g4f.ai. The credits section lists several community projects that contributed code, and the manifesto at the end lays out principles like open access to AI tooling, collaboration across providers, and opposition to closed, monopolistic systems.

• Source 1 · Project repository
• Source 2 · README