Hi Product Hunt 👋 I contribute to Gonka, an open-source network where independently owned GPUs serve open-weight LLMs behind an OpenAI-compatible API. The problem that kept hitting: people will not evaluate a new inference provider until they can test it for free. So the four independent brokers on the network each put up a welcome bonus - 100M tokens (DAHL), up to 100M (Gonka API), 1M (GG) and $20 in credits (Router). "Tokens" here means model tokens, not a cryptocurrency, and there is no wallet step. What you can run today: • DeepSeek V4-Flash — 380K–400K served context • GLM-5.3-Flash — 390K–400K • MiniMax M2.7 — 180K–200K It is a base-URL swap in the OpenAI SDK, Cline, Continue, LibreChat, SillyTavern or anything else that takes a custom endpoint. What it is not: the lowest-latency provider on the market. It suits long context, batch jobs and unattended agents rather than snappy chat. Each broker is an independent entity and sets its own limits and pricing after the bonus.

Gonka AI Drop

Free credits for open-weight LLMs on community GPUs

Free credits for open-weight LLMs on community GPUs