Llama 4 Scout
Llama 4 Scout is an open-weight language model from Meta, released on April 5, 2025. It is a mixture-of-experts model with 109B total parameters, of which 17B are active for each token. Meta lists a context window of 10M tokens. The model accepts text and image input.
- Developer
- Meta
- Released
- April 5, 2025
- Total parameters
- 109B
- Active parameters per token
- 17B
- Architecture
- MoE
- Context window
- 10M
- Input
- Text, image
- License
- Llama 4 Community License
- Commercial use
- Yes, with conditions
- Official weights, GB
- 217.3 GB
- Hugging Face repository
- huggingface.co
- Checked on
The weights are released under the Llama 4 Community License. Commercial use is allowed with conditions. Companies whose products had more than 700 million monthly active users on the release date need a separate license from Meta. Products must display “Built with Llama”, model names derived from it must start with “Llama”, and the Llama Acceptable Use Policy applies.
The official weights on Hugging Face take about 217 GB in BF16. Running the model needs at least that much memory across GPUs and system RAM, plus room for the context cache; quantized versions need less.
- Total parameters, billions
- 109B
- Active parameters, billions
- 17B
- Context window, tokens
- 10M
- License type
- Custom license
- License conditions
- Companies whose products had more than 700 million monthly active users on the release date need a separate license from Meta. Products must display "Built with Llama", model names derived from it must start with "Llama", and the Llama Acceptable Use Policy applies.
- Weights precision
- BF16
- License text
- github.com
- Sources
Page URL Hugging Face model page https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct Hugging Face API https://huggingface.co/api/models/meta-llama/Llama-4-Scout-17B-16E-Instruct?expand=safetensors