Gonka Broker — Cheap Decentralized LLM Inference with Simple Billing
Gonka Broker

Open-source LLM inference at a low cost

Gonka Broker is an API to a global decentralized GPU network, compatible with OpenAI and Anthropic clients. Sign up, add a card, get an API key. No wallets or crypto.

Developers choose Gonka Broker

We remove the crypto-infrastructure barriers. What's left is a plain API and cheap inference.

OpenAI and Anthropic APIs

Change the base URL in your OpenAI or Anthropic SDK. Works with Python, JavaScript, and any compatible client.

Billing in USD

Top up by card, get simple invoices. No tokens, wallets, or on-chain steps to manage.

API keys and accounts

Create separate keys per app and environment, with full access control from the dashboard.

Decentralized GPU network

Tap the power of Gonka's global GPU infrastructure for open-source model inference.

Start in three steps

Instant access without diving into the decentralized network.

1

Create an account

Sign up with email or Google. No wallets, seed phrases, or crypto.

2

Generate an API key

Create a key for your application or environment in the dashboard.

3

Send requests

Point your OpenAI or Anthropic client at our proxy and start inferring.

Works with your OpenAI or Anthropic client

Change two lines: the base URL and your API key. Your existing code, tools, and frameworks keep working.

from openai import OpenAI

client = OpenAI(
    base_url="https://proxy.gonkabroker.com/v1",
    api_key="gnk-prx-YOUR_KEY_HERE",
)

resp = client.chat.completions.create(
    model="MiniMaxAI/MiniMax-M2.7",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)

Using the Anthropic SDK instead? See the Anthropic-compatible setup.

Also works with Claude Code, Cursor, Continue, LibreChat, LangChain, LlamaIndex, and any tool with a custom OpenAI or Anthropic endpoint.

Free tokens every month

Every account includes a free tier of at least 1M tokens of any model, renewed every calendar month. No credit card. If you top up, your free tokens are still spent first.

About the free tier

Inference pricing

Run open-source models at rock-bottom token prices, with your USD rate locked in when you top up.

Model / Provider Input / 1M Output / 1M
OpenAI GPT-5 $1.25 $10.00
Claude 4.5 Sonnet $3.00 $15.00
Gonka · MiniMax M2 $0.30 $0.30
Gonka · Kimi K2 $0.35 $0.35
Gonka · BGE-M3 $0.01

Illustrative comparison. Prices change over time for every provider; always check current rates.

Accepted payment methods

Visa Mastercard Union Pay Apple Pay Google Pay Link and others
For developers

Works with your favourite tools

Point your OpenAI or Anthropic client at Gonka Broker and get instant access to open-source models at low cost. No custom SDKs, no rewrites.

  • Any OpenAI SDK (Python, JavaScript, Go): change base_url and api_key.
  • AI-powered tools: Claude Code, Cursor, Continue, LibreChat, and anything with a custom OpenAI or Anthropic endpoint.
  • Frameworks: LangChain, LlamaIndex, and other libraries using the OpenAI format.

Quick start

  1. 1 Sign up: an account in 30 seconds.
  2. 2 Create a key in the dashboard.
  3. 3 Set base_url to proxy.gonkabroker.com/v1 and ship.
Get a free API key

Frequently asked questions

Do I need crypto or a wallet to use Gonka?

No. Gonka Broker handles all on-chain work for you. You sign up with email, pay by card in USD, and get a standard API key.

Is the API really OpenAI-compatible?

Yes. Point any OpenAI SDK or tool at https://proxy.gonkabroker.com/v1 with your Gonka Broker key. No code rewrites.

Can I use Gonka Broker for free?

Yes. Every account includes a monthly free tier: at least 1M tokens of any model in the catalog, renewed on the 1st of each month. No credit card required; see the free LLM API page for details.

How is pricing calculated?

You pay per token in USD. The Gonka network is market-based; we set a fixed rate on top that's locked in when you top up, so your price won't change until you've used your balance. Input and output cost the same, and prices usually stay below typical centralized providers.

Which models are available?

Open-source models served on the Gonka network, including MiniMax M2 and Kimi K2, with GLM-5.2 coming soon. For search and RAG there's the BGE-M3 embedding model via /v1/embeddings. The catalog grows over time; check GET /v1/models.

Can I generate embeddings?

Yes. The BGE-M3 multilingual embedding model is served through the standard OpenAI-compatible /v1/embeddings endpoint, with the same base URL and API key as chat. Any OpenAI SDK or RAG framework works unchanged, and you're billed on input tokens only.

Which model should I use for RAG?

Use BGE-M3 for the retrieval side (it embeds your documents and queries into multilingual 1024-dimensional vectors) and pair it with Kimi K2 or MiniMax M2 for answer generation. Both steps go through the same API.

Can I use the Anthropic SDK or Claude Code?

Yes. The same key also works over our Anthropic-compatible API: point the Anthropic SDK at https://proxy.gonkabroker.com, or set ANTHROPIC_BASE_URL to run your Claude Code agent on open-source models like MiniMax M2 and Kimi K2.

Do you serve OpenAI or Anthropic models?

No. We serve open-source models (Kimi K2, MiniMax M2, BGE-M3) over both the OpenAI and Anthropic protocols. Gonka Broker is an independent service, not affiliated with OpenAI or Anthropic.

Do you store my prompts?

No. Gonka Broker doesn't store the content of your prompts or completions. We log only the metadata needed to bill you and prevent abuse: request and token counts, timestamps, and price.

We never store your prompts

Gonka Broker never logs or stores the content of your prompts or completions: not on disk, not in our logs, not even in transit. We keep only the metadata we need to bill you and prevent abuse: request and token counts, timestamps, and price.

Start using cheap inference now

No crypto wallets, no complexity. Just an API and cheap open-source LLMs.

Get started

Already have an account? Sign in