Free AI inference with OpenAI-compatible endpoints. No API keys to buy. No accounts to create. Connect a wallet, get a key, start building.
Everything you need to run AI inference without intermediaries.
No pay-per-token billing. No subscription fees. The network is sustained by contributors, not your wallet.
Works with any OpenAI-compatible client. Just change the base URL. No SDK changes needed.
OpenAI-compatible endpoints. Switch from any provider with a one-line base URL change.
All requests use HTTPS encryption. Your prompts and responses are protected in transit.
Optimized inference pipeline with streaming support. Token-by-token delivery starts immediately.
Every line of code is auditable. Fork it. Run your own network. Contribute back. It's all on GitHub.
From prompt to response in seconds, with zero cost.
Submit your request through the OpenAI-compatible API with your unique key.
Your gb_ API key is validated to ensure authorized access to the inference API.
The selected model processes your prompt and generates a response using optimized inference.
Output is streamed token-by-token back to your client via the OpenAI-compatible API.
Connect a wallet to generate a deterministic API key. No sign-up forms, no emails.
Choose a wallet to generate your unique API key.
—
—
curl -X POST https://api.xiaomimimo.com/v1/chat/completions ...
export MIMO_API_KEY=your_key
export MIMO_BASE_URL=https://api.xiaomimimo.com/v1
import requests
response = requests.post(
"https://api.xiaomimimo.com/v1/chat/completions",
headers={
"Authorization": "Bearer YOUR_MIMO_API_KEY",
"Content-Type": "application/json"
},
json={
"model": "mimo-v2.5-pro",
"messages": [{"role": "user", "content": "Hello!"}]
}
)
print(response.json()["choices"][0]["message"]["content"])
from openai import OpenAI
client = OpenAI(
api_key="YOUR_MIMO_API_KEY",
base_url="https://api.xiaomimimo.com/v1"
)
response = client.chat.completions.create(
model="mimo-v2.5-pro",
messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(
api_key="YOUR_MIMO_API_KEY",
base_url="https://api.xiaomimimo.com/v1",
model="mimo-v2.5-pro"
)
result = llm.invoke("Hello!")
print(result.content)
curl -X POST https://api.xiaomimimo.com/v1/chat/completions \
-H "Authorization: Bearer YOUR_MIMO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "mimo-v2.5-pro",
"messages": [{"role": "user", "content": "Hello!"}]
}'
const response = await fetch("https://api.xiaomimimo.com/v1/chat/completions", {
method: "POST",
headers: {
"Authorization": "Bearer YOUR_MIMO_API_KEY",
"Content-Type": "application/json"
},
body: JSON.stringify({
model: "mimo-v2.5-pro",
messages: [{ role: "user", content: "Hello!" }]
})
});
const data = await response.json();
console.log(data.choices[0].message.content);
# In your hermes-agent config (config.yaml):
providers:
mimo:
type: openai
api_key: YOUR_MIMO_API_KEY
base_url: https://api.xiaomimimo.com/v1
model: mimo-v2.5-pro
# Or via environment variables:
export HERMES_PROVIDER=openai
export OPENAI_API_KEY=YOUR_MIMO_API_KEY
export OPENAI_BASE_URL=https://api.xiaomimimo.com/v1
Use pip install openai or npm install openai to get started with the official SDK.
Export GITBLOCK_API_KEY as an environment variable or pass it directly to the client.
Set the base URL to https://api.xiaomimimo.com/v1 — that's the only change from OpenAI.
Send chat completions, embeddings, and more. Full OpenAI API compatibility.
Curated models optimized for different tasks. All accessible through a single API endpoint.
Xiaomi's most capable reasoning model. Top-tier performance on math, code, and complex logic tasks.
General-purpose MiMo model. Strong balance of speed and reasoning for everyday AI tasks.
Optimized for speed and efficiency. Ideal for latency-sensitive applications and high-throughput workloads.
Multimodal model supporting text and vision. Send images alongside text for visual understanding tasks.
The advantages of a truly open inference network.
No hidden fees, no rate limits behind paywalls. The network runs on community-contributed compute.
All requests use HTTPS encryption. No prompts are stored or used for training.
Reliable hosted infrastructure with automatic failover. Designed for consistent availability.
No geographic restrictions. Access the network from any country, any device, any time.
Phase 3 will add a decentralized node network. Community members will contribute GPU compute for shared access.
The entire codebase is open. Fork it, customize it, run your own private network.
Everything you need to know about GitBlock.