> For the complete documentation index, see [llms.txt](https://docs.nebulablock.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nebulablock.com/products/serverless-inference/vision.md).

# Vision

Send images to vision and multimodal models on Nebula Block and ask questions about them.

**Vision models** behave similarly to text models, but they can accept and interpret images as well.

## Models available

Two kinds of model can take images:

**Dedicated vision-language models**

| Model                  | Model ID                      | Context |
| ---------------------- | ----------------------------- | ------- |
| Qwen3-VL-Plus          | `Qwen/Qwen3-VL-Plus`          | 256K    |
| Qwen3-VL-Flash         | `Qwen/Qwen3-VL-Flash`         | 256K    |
| Qwen2.5-VL-7B-Instruct | `Qwen/Qwen2.5-VL-7B-Instruct` | 32K     |

**Multimodal chat models.** Most of the flagship chat models — Claude, GPT, Gemini, and others — accept images in the same `messages` array as text. See the [Model Catalog](/products/serverless-inference/model-catalog.md#text-generation-and-multimodal-chat); anything listed as multimodal works with the requests on this page.

## Using the Models

1. Sign in to the [console](https://console.nebulablock.com) and make sure you have credit.
2. Open [**Serverless**](https://console.nebulablock.com/serverless) and pick a vision or multimodal model.
3. Set your parameters, add your image and prompt, and send it.

The parameters to tweak are the same as in [Text Generation](/products/serverless-inference/text-generation.md), with the addition of the following:

* **Image**: The image to be processed by the model.

### Through API Endpoint

This option is to use our API endpoint directly in your projects. Below are some code snippets to get you started!

> **NOTE:** Don't forget to use **your** API key. See the [API Reference](/api-reference/authentication.md) and the [Overview](/account/api-keys.md) for more details on authentication.

#### Using cURL

```bash
curl -X POST "https://inference.nebulablock.com/v1/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $NEBULA_API_KEY" \
    --data-raw '{
        "messages": [
			{"role":"user","content":[
			{"type":"image_url","image_url":
			{"url":"https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"}},
			{"type":"text","text":"What is this image?"}
		]}],
        "model": "Qwen/Qwen2.5-VL-7B-Instruct",
        "max_tokens": null, 
        "temperature": 1,
        "top_p": 0.9,
        "stream": false
    }'
```

#### Using Python

```python
import requests 
import os
 
url = "https://inference.nebulablock.com/v1/chat/completions"

headers = { 
    "Content-Type": "application/json", 
    "Authorization": f"Bearer {os.environ.get('NEBULA_API_KEY')}" 
} 
 
data = {
    "messages":[
		{"role":"user","content":[
		{"type":"image_url","image_url":
		{"url":"https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"}},
		{"type":"text","text":"What is this image?"}
	]}],
    "model":"Qwen/Qwen2.5-VL-7B-Instruct",
    "max_tokens":None,
    "temperature":1,
    "top_p":0.9,
    "stream":False
}

response = requests.post(url, headers=headers, json=data) 
print(response.json())
```

#### Using JavaScript

```javascript
const url = "https://inference.nebulablock.com/v1/chat/completions";

const headers = {
    "Content-Type": "application/json",
    "Authorization": `Bearer sk-kmlKFCXGBZl-PS0MLnpzAw`
};

const data = {
    messages: [
        {"role":"user","content":[
        {"type":"image_url","image_url":
        {"url":"https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"}},
        {"type":"text","text":"What is this image?"}
    ]}],
    model: 'Qwen/Qwen2.5-VL-7B-Instruct',
    max_tokens: null,
    temperature: 1,
    top_p: 0.9,
    stream: false
};

fetch(url, {
    method: 'POST',
    headers: headers,
    body: JSON.stringify(data)
})
    .then(response => response.json())
    .then(data => {
        console.log(JSON.stringify(data, null, 2));
    })
    .catch(error => console.error('Error:', error));
```

#### Selecting a Model

To specify the desired model, use this mapping for the `model_name`:

* Qwen2.5-VL-7B-Instruct: `Qwen/Qwen2.5-VL-7B-Instruct`

#### Response Example

A successful generation response (non-streaming) will contain a `chat.completion` object, and should look like this:

```json
    {
    "id": "chatcmpl-7ba48f119a564f4ea02b6a41386a3e40",
    "created": 1740689977,
    "model": "Qwen/Qwen2.5-VL-7B-Instruct",
    "object": "chat.completion",
    "system_fingerprint": null,
    "choices": [
        {
            "finish_reason": "stop",
            "index": 0,
            "message": {
                "content": "This image shows ....",
                "role": "assistant",
                "tool_calls": null,
                "function_call": null
            }
        }
    ],
    "usage": {
        "completion_tokens": 91,
        "prompt_tokens": 3604,
        "total_tokens": 3695,
        "completion_tokens_details": null,
        "prompt_tokens_details": null
    },
    "service_tier": null,
    "prompt_logprobs": null
}
```

As is the case with text generation, this represents the entire generated response, and setting the `stream` parameter to `True` will return the response in chunks:

```json
{
    "id": "chatcmpl-3812731562554b23a32dd80fbb7d0d09",
    "created": 1740692435,
    "model": "Qwen/Qwen2.5-VL-7B-Instruct",
    "object": "chat.completion.chunk",
    "choices": [
        {
            "index": 0,
            "delta": {
                "content": " setting"
            }
        }
    ]
}
{ 
  ...
}
...
```

with the collection of chunks forming the entire response.

Feel free to explore refer to the [API Reference](/api-reference/inference-api/vision.md) for more details.