
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.
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"}]
}'可用模型
資料中心覆蓋範圍
可用性目標

從你需要的部分開始。每一項都連到對應的參考文件
在單一個相容 OpenAI 的端點上執行對話、視覺與推理
Python 與 TypeScript 用戶端。可直接替換 OpenAI SDK
在受管運算資源上部署並執行 agent
快速上手指南、參考文件與程式碼範例
建立金鑰、挑一個模型、讀取串流回應
在 console 中產生 API 金鑰
複製 model id,或用 GET /v1/models 列出所有模型
呼叫端點並讀取串流回應
改用專用端點以取得穩定的延遲表現
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="")同一組 base URL 與金鑰,各路由的請求格式完全一致

這些應用運行在同一組 API 與運算資源上。開啟其中一個看看實際運作

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.

Your Spotify year, sorted into chapters. Claude names your eras, roasts your top songs, and recommends what's missing. Gemini paints each chapter.

Photorealistic image compositing. Gemini 3.1 Pro reads the scene's light and perspective; Nano Banana 2 paints your objects in subtly and naturally.
單一 API 背後有 200+ 個開源與前沿模型。把 /v1/models 中的任一 id 當作 model 欄位傳入即可,不需重新部署,也不用換 SDK
anthropic/claude-opus-4.8
Anthropic flagship
openai/gpt-5.6
OpenAI flagship
google/gemini-3.1-pro-preview
Text + vision
openai/sora-2
Text + image to video
moonshotai/kimi-k3
Open flagship, text-to-image
moonshotai/Kimi-K2-Thinking
Long-horizon agentic
價格依模型而定,可用模型請見「模型」區塊

建立金鑰、複製程式碼片段,就能收到回傳的 token