IBM Llama 4 Scout
IBM Llama 4 Scout
IBM Llama 4 Scout is an auto-regressive language model that uses a mixture-of-experts (MoE) architecture and incorporates early fusion for native multimodality.
Model details
| Item | Value | Description |
|---|---|---|
| Model name | IBM Llama 4 Scout | The name of the model. |
| API model name | ibm__llama_4_scout | The name of the model that is used in the Box AI API for model overrides. The user must provide this exact name for the API to work. |
| Hosting layer | IBM | The trusted organization that securely hosts LLM. |
| Model provider | Meta | The organization that provides this model. |
| Release date | April 5th 2025 | The release date for the model. |
| Knowledge cutoff date | August 2024 | The date after which the model does not get any information updates. |
| Input context window | 10m | The number of tokens supported by the input context window. |
| Maximum output tokens | Not specified | The number of tokens that can be generated by the model in a single request. |
| Empirical throughput | Not specified | The number of tokens the model can generate per second. |
| Open source | Yes | Specifies if the model's code is available for public use. |
| IP infringement protection | No | Use of this model does not come with any intellectual property rights assurances or protections from Box. Please consider any potential IP issues that might arise from using the model’s outputs. |
Additional documentation
For additional information, see official IBM Llama 4 Scout documentation.