> ## Documentation Index
> Fetch the complete documentation index at: https://docs.modelslab.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Selection Guide

> How to discover, choose, and use the right AI model for your generation requests across image, video, audio, and LLM endpoints

# Model Selection Guide

Every generation request on ModelsLab requires a `model_id` parameter that tells the API which AI model to use. With 50,000+ models available, this guide helps you discover and choose the right one.

***

## How model\_id Works

The `model_id` is a string identifier you pass in your API request body. It determines which model processes your generation.

```python theme={null}
import requests

response = requests.post(
    "https://modelslab.com/api/v7/images/text-to-image",
    json={
        "key": "your_api_key",
        "model_id": "flux",        # <-- this selects the model
        "prompt": "a red apple on a white table",
        "width": 512,
        "height": 512
    }
)
```

```bash theme={null}
curl -X POST "https://modelslab.com/api/v7/images/text-to-image" \
  -H "Content-Type: application/json" \
  -d '{
    "key": "your_api_key",
    "model_id": "flux",
    "prompt": "a red apple on a white table",
    "width": 512,
    "height": 512
  }'
```

***

## Three Ways to Discover Models

### 1. API: Models Endpoint

Query the models API to search programmatically:

```bash theme={null}
# Search by name
curl "https://modelslab.com/api/v7/models?search=flux&key=your_api_key"

# Filter by feature
curl "https://modelslab.com/api/v7/models?feature=imagen&key=your_api_key"
```

### 2. CLI: modelslab models

The CLI provides rich model discovery commands:

```bash theme={null}
# Search by name
modelslab models search --search "flux"

# Filter by feature category
modelslab models search --feature imagen         # Image models
modelslab models search --feature video_fusion   # Video models
modelslab models search --feature audio_gen      # Audio models
modelslab models search --feature llmaster       # LLM/chat models

# Get full details about a model
modelslab models detail --id flux

# JSON output for scripting
modelslab models search --search "flux" --output json --jq '.[].model_id'
```

### 3. Web: Model Browser

Browse and filter all models visually at [modelslab.com/models](https://modelslab.com/models).

***

## Models by Category

### Image Generation

Use with `/api/v7/images/text-to-image`, `/api/v7/images/image-to-image`, and `/api/v7/images/inpaint`.

| model\_id    | Name                | Best For                          |
| ------------ | ------------------- | --------------------------------- |
| `flux`       | Flux Dev            | Fast, high-quality general images |
| `midjourney` | MidJourney          | Artistic, stylized images         |
| `sdxl`       | Stable Diffusion XL | Versatile base model              |
| `imagen-3`   | Google Imagen 3     | Premium photorealistic images     |
| `imagen-4`   | Google Imagen 4     | Latest Google image model         |

<Note>
  For image-to-image and inpainting with SD/SDXL models (e.g., `midjourney`), include the `scheduler` parameter. Flux models don't require it.
</Note>

### Video Generation

Use with `/api/v7/video-fusion/text-to-video` and `/api/v7/video-fusion/image-to-video`.

| model\_id      | Name                    | Type                |
| -------------- | ----------------------- | ------------------- |
| `seedance-t2v` | Seedance Text-to-Video  | Text to video       |
| `seedance-i2v` | Seedance Image-to-Video | Image to video      |
| `wan2.2`       | Wan 2.1                 | Text/image to video |
| `wan2.6-t2v`   | Wan 2.6 Text-to-Video   | Text to video       |
| `wan2.6-i2v`   | Wan 2.6 Image-to-Video  | Image to video      |
| `veo2`         | Google Veo 2            | Premium video       |
| `veo3`         | Google Veo 3            | Latest Google video |
| `sora-2`       | OpenAI Sora 2           | Premium video       |

<Note>
  Video generation is asynchronous. The API returns a `processing` status with an `id` — use `/api/v7/video-fusion/fetch/{id}` to poll for results.
</Note>

### Audio & Voice

Use with `/api/v7/voice/*` endpoints.

| model\_id                | Name                       | Endpoint         |
| ------------------------ | -------------------------- | ---------------- |
| `eleven_multilingual_v2` | ElevenLabs Multilingual v2 | text-to-speech   |
| `eleven_english_sts_v2`  | ElevenLabs Voice Changer   | speech-to-speech |
| `scribe_v1`              | ElevenLabs Scribe          | speech-to-text   |
| `eleven_sound_effect`    | ElevenLabs SFX             | sound-generation |
| `music_v1`               | ElevenLabs Music           | music-gen        |
| `inworld-tts-1`          | Inworld TTS                | text-to-speech   |

<Warning>
  Text-to-speech uses the `prompt` parameter (not `text`) and requires a valid ElevenLabs `voice_id` (e.g., `21m00Tcm4TlvDq8ikWAM` for Rachel).
</Warning>

### LLM / Chat

Use with `/api/v7/llm/chat/completions` (OpenAI-compatible format).

| model\_id                                       | Name                 | Provider |
| ----------------------------------------------- | -------------------- | -------- |
| `meta-llama-3-8B-instruct`                      | Llama 3 8B Instruct  | Meta     |
| `meta-llama-Llama-3.3-70B-Instruct-Turbo`       | Llama 3.3 70B Turbo  | Meta     |
| `meta-llama-Meta-Llama-3.1-405B-Instruct-Turbo` | Llama 3.1 405B Turbo | Meta     |
| `deepseek-ai-DeepSeek-R1-Distill-Llama-70B`     | DeepSeek R1          | DeepSeek |
| `deepseek-ai-DeepSeek-V3`                       | DeepSeek V3          | DeepSeek |
| `gemini-2.0-flash-001`                          | Gemini 2.0 Flash     | Google   |
| `gemini-2.5-pro`                                | Gemini 2.5 Pro       | Google   |
| `Qwen-Qwen2.5-72B-Instruct-Turbo`               | Qwen 2.5 72B Turbo   | Qwen     |
| `mistralai-Mixtral-8x7B-Instruct-v0.1`          | Mixtral 8x7B         | Mistral  |

<Note>
  The chat completions endpoint accepts both `model_id` and `model` (OpenAI-compatible alias). The `messages` array follows the standard OpenAI format.
</Note>

***

## Code Examples

### Image Generation (Python)

```python theme={null}
import requests

response = requests.post(
    "https://modelslab.com/api/v7/images/text-to-image",
    json={
        "key": "your_api_key",
        "model_id": "flux",
        "prompt": "a futuristic cityscape at sunset, highly detailed",
        "width": 1024,
        "height": 1024,
        "samples": 1
    }
)

data = response.json()
if data["status"] == "success":
    print(f"Image URL: {data['output'][0]}")
```

### Chat Completion (Python)

```python theme={null}
response = requests.post(
    "https://modelslab.com/api/v7/llm/chat/completions",
    json={
        "key": "your_api_key",
        "model_id": "meta-llama-3-8B-instruct",
        "messages": [
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": "Explain quantum computing briefly."}
        ],
        "max_tokens": 200,
        "temperature": 0.7
    }
)

data = response.json()
print(data["choices"][0]["message"]["content"])
```

### Chat Completion (OpenAI SDK Compatible)

```python theme={null}
from openai import OpenAI

client = OpenAI(
    base_url="https://modelslab.com/api/v7/llm",
    api_key="your_api_key"
)

response = client.chat.completions.create(
    model="meta-llama-3-8B-instruct",
    messages=[{"role": "user", "content": "Hello!"}],
    max_tokens=50
)

print(response.choices[0].message.content)
```

### CLI Examples

```bash theme={null}
# Generate image with specific model
modelslab generate image --prompt "sunset over mountains" --model flux

# Chat with an LLM
modelslab generate chat --message "Explain AI" --model meta-llama-3-8B-instruct

# Generate video
modelslab generate video --prompt "ocean waves" --model seedance-t2v

# Set a default model
modelslab config set generation.default_model flux
```

***

## Tips for Choosing the Right Model

1. **Start with popular models** — `flux` for images, `meta-llama-3-8B-instruct` for chat
2. **Use feature filters** — `modelslab models search --feature imagen` narrows results
3. **Check model details** — `modelslab models detail --id <model_id>` shows supported parameters
4. **Consider provider** — models from different providers (Google, Meta, ElevenLabs) have different capabilities and pricing
5. **Test with small requests first** — use low resolution/token counts while experimenting

***

## Resources

* [Model Browser](https://modelslab.com/models) — Visual model discovery
* [API Reference](/api-reference) — Full endpoint documentation
* [CLI Installation](https://github.com/ModelsLab/modelslab-cli) — Install the CLI tool