Gonka API pricing
The Gonka network is market-based — we turn that into something predictable: a fixed USD price per token, locked in when you top up. You see exactly how many tokens you'll get, input and output cost the same, and it stays below typical centralized providers.
Current token prices
Per 1M tokens, in USD. Centralized providers shown for reference.
| Model | Input / 1M | Output / 1M |
|---|---|---|
| Qwen3-235B-A22B-Instruct | $0.35 | $0.35 |
| Kimi-K2 | $0.35 | $0.35 |
| — OpenAI GPT-5 (for reference) | $1.25 | $10.00 |
| — Claude 4.5 Sonnet (for reference) | $3.00 | $15.00 |
The Gonka network's prices move with GPU supply and demand; we set a fixed
USD rate on top of it. If we change that rate, it applies only to future
top-ups — the balance you've already paid for keeps its price until you use
it. The current model list and rates are always in your dashboard and via
GET /v1/models.
Pricing you can plan around
Open-source inference in plain USD — here's what makes the price work in your favour.
Locked in when you top up
Your per-token price is fixed the moment you pay, for your entire balance. At checkout you see exactly how many tokens of each model you'll get — and the rate won't change until you've used them.
Same price for input and output
We charge the same per token whether it's your prompt or the model's reply. Centralized providers usually charge several times more for output, so generation-heavy workloads cost far less here.
Below centralized providers
Run open-source models at a USD price that undercuts the big centralized APIs — without touching crypto, wallets, or tokens.
Accepted payment methods
and others
See your rate in the dashboard
Create an account to view current per-token prices and start sending requests.
Sign up and startAlready have an account? Sign in