A free ChatGPT API proxy with a business model shaped like a donut

What it does

free_chatgpt_api is a hosted API proxy at https://free.v36.cm that speaks standard OpenAI protocol. You grab a free key, point your existing client at their base URL, and GPT-4o-mini or GPT-3.5-turbo just works. The repo is essentially documentation and screenshots showing how to wire it into ChatGPT-Next-Web, Lobe Chat, FastGPT, and a dozen others.

The interesting bit

The free tier is deliberately capped—96 RPM, no support, no GPT-4, and the maintainers openly complain about bot farms burning keys. The “interesting bit” is the honesty: the README spends as much space upselling the paid API (api.v36.cm) as explaining the free one. It is a lead-gen funnel wearing the skin of a public-benefit project.

Key highlights

• Free models: gpt-4o-mini, gpt-3.5-turbo variants; GPT-4 is “periodically limited” and text-davinci is gone
• Claims direct China-accessible connection without proxy
• Paid tier promises 130+ models, Midjourney, Suno, and “2.8–3×” discount vs. OpenAI pricing
• Heavy legal disclaimers: non-commercial only, no unregistered generative services to Chinese public
• Star history graph at the bottom suggests the repo tracks popularity more than code

Caveats

• No source code in the repository; it is a landing page for a third-party service
• Free tier explicitly “does not guarantee stability” and offers zero technical support
• Rate limiting and CC-level bot protection suggest the free keys are already under strain

Verdict

Worth a bookmark if you need a quick, zero-setup OpenAI drop-in for personal tinkering and can tolerate the RPM ceiling. Skip it if you are building anything that needs reliability, or if you are uncomfortable with a black-box proxy handling your prompts and API keys.