llama4

The Llama 4 collection of models are natively multimodal AI models that enable text and multimodal experiences. These two models leverage a mixture-of-experts (MoE) architecture and support native multimodality (image input).

Supported languages: Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai, and Vietnamese.

Input: multilingual text, image

Output: multilingual text, code

Models

Llama 4 Scout

ollama run llama4:scout

109B parameter MoE model with 17B active parameters

Llama 4 Maverick

ollama run llama4:maverick

400B parameter MoE model with 17B active parameters

Intended Use

Intended Use Cases: Llama 4 is intended for commercial and research use in multiple languages. Instruction tuned models are intended for assistant-like chat and visual reasoning tasks, whereas pretrained models can be adapted for natural language generation. For vision, Llama 4 models are also optimized for visual recognition, image reasoning, captioning, and answering general questions about an image. The Llama 4 model collection also supports the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. The Llama 4 Community License allows for these use cases.

Out-of-scope: Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in any other way that is prohibited by the Acceptable Use Policy and Llama 4 Community License. Use in languages or capabilities beyond those explicitly referenced as supported in this model card.

Note:

Llama 4 has been trained on a broader collection of languages than the 12 supported languages (pre-training includes 200 total languages). Developers may fine-tune Llama 4 models for languages beyond the 12 supported languages provided they comply with the Llama 4 Community License and the Acceptable Use Policy. Developers are responsible for ensuring that their use of Llama 4 in additional languages is done in a safe and responsible manner.
Llama 4 has been tested for image understanding up to 5 input images. If leveraging additional image understanding capabilities beyond this, Developers are responsible for ensuring that their deployments are mitigated for risks and should perform additional testing and tuning tailored to their specific applications.

Benchmarks

Category	Benchmark	# Shots	Metric	Llama 3.3 70B	Llama 3.1 405B	Llama 4 Scout	Llama 4 Maverick
Image Reasoning	MMMU	0	accuracy	No multimodal support		69.4	73.4
	MMMU Pro^	0	accuracy			52.2	59.6
	MathVista	0	accuracy			70.7	73.7
Image Understanding	ChartQA	0	relaxed_accuracy			88.8	90.0
	DocVQA (test)	0	anls			94.4	94.4
Code	LiveCodeBench (10/01/2024-02/01/2025)	0	pass@1	33.3	27.7	32.8	43.4
Reasoning & Knowledge	MMLU Pro	0	macro_avg/acc	68.9	73.4	74.3	80.5
	GPQA Diamond	0	accuracy	50.5	49.0	57.2	69.8
Multilingual	MGSM	0	average/em	91.1	91.6	90.6	92.3
Long Context	MTOB (half book) eng->kgv/kgv->eng	-	chrF	Context window is 128K		42.2 / 36.6	54.0 / 46.4
	MTOB (full book) eng->kgv/kgv->eng	-	chrF			39.7 / 36.3	50.8 / 46.7

*reported numbers for MMMU Pro is the average of Standard and Vision tasks

Reference

Meta Llama 4 post

![image.png](/assets/library/llama4/e7b65449-da66-4a50-b36c-d0c2359295b4)

**Supported languages:** Arabic, English, French, German, Hindi, Indonesian, Italian, Portuguese, Spanish, Tagalog, Thai, and Vietnamese.

**Input:** multilingual text, image

**Output:** multilingual text, code

### Models

#### Llama 4 Scout

```
ollama run llama4:scout
```

109B parameter MoE model with 17B active parameters

#### Llama 4 Maverick

```
ollama run llama4:maverick
```

400B parameter MoE model with 17B active parameters

### Intended Use

**Intended Use Cases:** Llama 4 is intended for commercial and research use in multiple languages. Instruction tuned models are intended for assistant-like chat and visual reasoning tasks, whereas pretrained models can be adapted for natural language generation. For vision, Llama 4 models are also optimized for visual recognition, image reasoning, captioning, and answering general questions about an image. The Llama 4 model collection also supports the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. The Llama 4 Community License allows for these use cases.

**Out-of-scope**: Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in any other way that is prohibited by the Acceptable Use Policy and Llama 4 Community License. Use in languages or capabilities beyond those explicitly referenced as supported in this model card.

Note:

1. Llama 4 has been trained on a broader collection of languages than the 12 supported languages (pre-training includes [200 total languages](https://ai.meta.com/research/no-language-left-behind/)). Developers may fine-tune Llama 4 models for languages beyond the 12 supported languages provided they comply with the Llama 4 Community License and the Acceptable Use Policy.  Developers are responsible for ensuring that their use of Llama 4 in additional languages is done in a safe and responsible manner.

2. Llama 4 has been tested for image understanding up to 5 input images. If leveraging additional image understanding capabilities beyond this, Developers are responsible for ensuring that their deployments are mitigated for risks and should perform additional testing and tuning tailored to their specific applications.

### Benchmarks

<table>
  <tr>
    <th>Category</th>
    <th>Benchmark</th>
    <th># Shots</th>
    <th>Metric</th>
    <th>Llama 3.3 70B</th>
    <th>Llama 3.1 405B</th>
    <th>Llama 4 Scout</th>
    <th>Llama 4 Maverick</th>
  </tr>
  <tr>
    <td> Image Reasoning</td>
    <td>MMMU</td>
    <td>0</td>
    <td>accuracy</td>
    <td colspan="2"> No multimodal support </td>
    <td>69.4</td>
    <td>73.4</td>
  </tr>
  <tr>
    <td>  </td>
    <td>MMMU Pro^</td>
    <td>0</td>
    <td>accuracy</td>
    <td colspan="2"> </td>
    <td>52.2</td>
    <td>59.6</td>
  </tr>
  <tr>
    <td>  </td>
    <td>MathVista</td>
    <td>0</td>
    <td>accuracy</td>
    <td colspan="2">  </td>
    <td>70.7</td>
    <td>73.7</td>
  </tr>
  <tr>
    <td> Image Understanding </td>
    <td>ChartQA</td>
    <td>0</td>
    <td>relaxed_accuracy</td>
    <td colspan="2">  </td>
    <td>88.8</td>
    <td>90.0</td>
  </tr>
  <tr>
    <td>  </td>
    <td>DocVQA (test)</td>
    <td>0</td>
    <td>anls</td>
    <td colspan="2"> </td>
    <td>94.4</td>
    <td>94.4</td>
  </tr>
  <tr>
    <td> Code  </td>
    <td> LiveCodeBench
(10/01/2024-02/01/2025)
 </td>
    <td>0</td>
    <td>pass@1</td>
    <td> 33.3 </td>
    <td> 27.7 </td>
    <td>32.8</td>
    <td>43.4</td>
  </tr>
  <tr>
    <td> Reasoning & Knowledge  </td>
    <td> MMLU Pro</td>
    <td>0</td>
    <td>macro_avg/acc</td>
    <td> 68.9 </td>
    <td> 73.4 </td>
    <td>74.3</td>
    <td>80.5</td>
  </tr>
    <tr>
    <td>  </td>
    <td> GPQA Diamond</td>
    <td>0</td>
    <td>accuracy</td>
    <td> 50.5 </td>
    <td> 49.0 </td>
    <td>57.2</td>
    <td>69.8</td>
  </tr>
    <tr>
    <td> Multilingual  </td>
    <td> MGSM</td>
    <td>0</td>
    <td>average/em</td>
    <td> 91.1 </td>
    <td> 91.6</td>
    <td>90.6</td>
    <td>92.3</td>
  </tr>
    <tr>
    <td> Long Context  </td>
    <td> MTOB (half book) eng->kgv/kgv->eng </td>
    <td>-</td>
    <td>chrF</td>
    <td colspan="2"> Context window is 128K </td>
    <td>42.2 / 36.6</td>
    <td>54.0 / 46.4</td>
  </tr>
    <tr>
    <td> </td>
    <td> MTOB (full book) eng->kgv/kgv->eng </td>
    <td>-</td>
    <td>chrF</td>
    <td colspan="2">    </td>
    <td>39.7 / 36.3</td>
    <td>50.8 / 46.7</td>
  </tr>
</table>
*reported numbers for MMMU Pro is the average of Standard and Vision tasks

### Reference
- Meta Llama 4 [post](https://ai.meta.com/blog/llama-4-multimodal-intelligence/)

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)