DeepSeek-V3

LLM SGLang H200

A state-of-the-art 671B-parameter MoE LLM licensed for commercial use

Deploy DeepSeek-V3 behind an API endpoint in seconds.

Talk to sales

Example usage

DeepSeek is optimized using SGLang and uses an OpenAI-compatible API endpoint.

Input

1import httpx
2import os
3
4MODEL_ID = "abcd1234"  # Replace with your model ID
5DEPLOYMENT_ID = "abcd1234"  # [Optional] Replace with your deployment ID
6API_KEY = os.environ["BASETEN_API_KEY"]
7
8resp = httpx.post(
9    f"https://model-{MODEL_ID}.api.baseten.co/environments/production/predict",
10    headers={"Authorization": f"Api-Key {API_KEY}"},
11    json={
12        "model": "deepseek_v3",
13        "messages": [
14            {"role": "system", "content": "You are a helpful AI assistant"},
15            {"role": "user", "content": "Write FizzBuzz in Python"},
16        ],
17        "max_tokens": 1024,
18    },
19    timeout=None
20)
21
22print(resp.json())

JSON output

1{
2    "id": "8456fe51db3548789f199cfb8c8efd35",
3    "object": "text_completion",
4    "created": 1735236968,
5    "model": "/models/deepseek_v3",
6    "choices": [
7        {
8            "index": 0,
9            "text": "FizzBuzz is a classic programming problem where you print numbers from 1 to 100...",
10            "logprobs": null,
11            "finish_reason": "stop",
12            "matched_stop": 1
13        }
14    ],
15    "usage": {
16        "prompt_tokens": 14,
17        "total_tokens": 240,
18        "completion_tokens": 226,
19        "prompt_tokens_details": null
20    }
21}

Example usage

Deploy any model in just a few commands