Baseten acquires Parsed: Own your intelligence by unifying training and inference. READ

large language

Qwen3 Omni Thinker

An "omni" model that can process both image and audio input

Deploy now

‌

Model details

Developed by
Qwen
Model family
Qwen
Use case
large language
Version
Omni
Variant
Thinker
Size
30B
API
openai-compatible
License
Apache 2.0

View repository

Example usage

Qwen 3 Omni is compatible with the OpenAI SDK. It takes multiple modalities of input: text, image, and audio. The "Thinker" variant of the model, implemented here, returns text.

1{
2  "model": "qwen3-omni",
3  "messages": [
4    {"role": "system", "content": "You are a helpful assistant."},
5    {
6      "role": "user",
7      "content": [
8        {"type": "text", "text": "Describe what you see and hear."},
9        {
10          "type": "image_url",
11          "image_url": {"url": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-Omni/demo/cars.jpg"}
12        },
13        {
14          "type": "audio_url",
15          "audio_url": {"url": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-Omni/demo/cough.wav"}
16        }
17      ]
18    }
19  ],
20  "max_tokens": 2048,
21  "temperature": 0.7,
22  "stream": false
23}

Input

1from openai import OpenAI
2import os
3
4client = OpenAI(
5    api_key=os.environ["BASETEN_API_KEY"],
6    base_url="https://model-xxxxxx.api.baseten.co/environments/production/sync/v1"
7)
8
9resp = client.chat.completions.create(
10    model="qwen3-omni",
11    messages=[
12        {"role": "system", "content": "You are a helpful assistant."},
13        {"role": "user", "content": [
14            {"type": "text", "text": "Describe this image and audio content."},
15            {"type": "image_url", "image_url": {"url": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-Omni/demo/cars.jpg"}},
16            {"type": "audio_url", "audio_url": {"url": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-Omni/demo/cough.wav"}}
17        ]}
18    ],
19    max_tokens=2048,
20    temperature=0.7,
21    stream=False,
22)
23print(resp.choices[0].message.content)

JSON output

1{
2    "id": "chatcmpl-...",
3    "object": "chat.completion",
4    "created": 1710000000,
5    "model": "qwen3-omni",
6    "choices": [
7        {
8            "index": 0,
9            "finish_reason": "stop",
10            "message": {
11                "role": "assistant",
12                "content": "I see several parked cars in front of a building and hear a short cough."
13            }
14        }
15    ],
16    "usage": {
17        "prompt_tokens": 512,
18        "completion_tokens": 24,
19        "total_tokens": 536
20    }
21}