Gonka Pricing
Per-token pricing for the 1 open-weight LLMs on Gonka — a decentralized inference network that pools GPUs into one OpenAI-compatible API, accessed through third-party brokers.
Snapshot as of: Jun 28, 2026
Gonka free tier & credits
No ongoing free tier — see credits & cheapest models — see the full Gonka free tier guide
How Gonka pricing works: Gonka charges a single blended per-token rate — input and output are priced identically. The rate is network-subsidized and recalculated every block based on utilization, so it rises as the network gets busier. The figures below are a snapshot from a public broker (GonkaGate), not a fixed price.
Author | Model | Context | Input/1M | Output/1M |
|---|---|---|---|---|
MM | 180k | $0.000307 | $0.000307 |
Blended per-token rate (input and output priced identically), per 1M tokens. Snapshot from the GonkaGate broker — Gonka's network price is dynamic and recalculated per block.
About Gonka
Gonka is a decentralized network for AI inference: instead of one data center, thousands of GPUs around the world are pooled into a single OpenAI-compatible API. It is an open-source project incubated by Product Science and backed by Bitfury, and it serves leading open-weight models such as Kimi, MiniMax and Qwen.
You don't connect to Gonka directly — access is resold by third-party brokers (GonkaGate, JoinGonka, OpenGNK and others), each of which sets its own fees, payment options and rate limits on top of the network cost. Prices are extremely low today because the network is underutilized and partly subsidized; they are recalculated every block and will rise as usage grows.
Gonka Pricing FAQ
How much does Gonka cost?
Gonka charges a single blended per-token rate — input and output are priced the same. At the most recent snapshot that rate was about $0.000314 per million tokens via the GonkaGate broker. The rate is subsidized and recalculated every block, so it rises as the network gets busier.
Why is Gonka so cheap?
Two reasons: the network is currently underutilized, and its dynamic-pricing algorithm lowers prices when utilization is low to attract usage. An initial zero-cost grace period has already ended; today’s price is the live demand-based rate, not free credits.
Which models does Gonka serve?
A small set of leading open-weight models — Kimi K2.6, MiniMax M2.7, and Qwen3 235B A22B Instruct. Use GET /v1/models against your broker to confirm current model IDs and availability.
Is Gonka OpenAI-compatible?
Yes. You point the OpenAI SDK at a broker base URL (e.g. https://<broker>/v1) with an API key and call /v1/chat/completions as usual — no Gonka-specific client needed.
Is the price fixed?
No. Gonka uses dynamic pricing recalculated every block based on network utilization. The figures here are a manually-captured snapshot from one broker and will drift over time.