One endpoint,
200+ models

Keep your code, change the model ID. Chat, vision, and reasoning on H200 GPUs

curl https://api.gmi-serving.com/v1/chat/completions \
  -H "Authorization: Bearer $GMI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-V4-Pro",
    "stream": true,
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Models Available: 200+
Data Center Coverage: Global
Uptime target: 99.9%

Four ways in

Start with the piece you need. Each links to its reference

Inference API

Chat, vision, and reasoning on one OpenAI-compatible endpoint

API Reference

SDK

Python and TypeScript clients. Drop-in for the OpenAI SDK

SDK Docs

AgentBox

Deploy and run agents on managed compute

Open AgentBox

Docs

Quickstarts, references, and code samples

Read Docs

First request

Create a key, pick a model, read the stream

01
Create a key
Generate an API key in the console
02
Pick a model
Copy a model id, or list them with GET /v1/models
03
Send a request
Call the endpoint and read the stream
04
Scale
Switch to a dedicated endpoint for fixed latency

python · openai client

# pip install openai
from openai import OpenAI
client = OpenAI(base_url="https://api.gmi-serving.com/v1",
                api_key=GMI_API_KEY)  # from console.gmicloud.ai

stream = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V4-Pro",  # any id from /v1/models
    stream=True,
    messages=[{"role":"user",
               "content":"Summarize this paragraph."}])

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

Endpoints

One base URL, one key, the same request shape across routes

Base URLhttps://api.gmi-serving.com

POST/v1/chat/completionsChat, vision, tools, streamed or full

POST/v1/responsesOpenAI Responses API

POST/v1/messagesAnthropic Messages API

GET/v1/modelsList models on your key

Built on GMI Cloud

Apps running on the same API and compute. Open one to see it work

See all demos

DeepSeek V4 Pro, GPT 5.5

Model Arena

Compare top LLMs like DeepSeek V4 Pro and GPT 5.5 side by side on real coding, math, and system design prompts, benchmarked live on GMI Cloud's inference API for speed, cost, and output quality.

open demo