Model · Meta · Text
Llama 4 Scout
Meta's small/efficient Llama 4 variant — natively multimodal MoE architecture. The practical Llama 4 entry point for self-hosted multimodal applications.
- Modality
- Text
- License
- Llama 4 Community License (Open weights)
- Parameter size
- 17B active / 16E
- Released
- April 5, 2025
- Last verified
- September 8, 2026
- Runs locally
- Yes
- Also handles
- vision
What you need to run it
- Mac
- Yes. Any Apple Silicon Mac (M1 or newer) with at least 96 GB (128 GB recommended) of memory. Newer chips are faster, but memory is what decides whether it runs.
- Windows / Linux PC
- NVIDIA works best. A graphics card with at least 80 GB of video memory.
- No graphics card
- Not practical. Too slow without a graphics card or Apple Silicon.
Fits a 96 GB or 128 GB Mac and runs well in MLX, Ollama and LM Studio because only 17B of it is active per word.
Numbers are for: 4-bit build of the 109B model (about 65 GB). Our estimate from the model’s size — the maker publishes no figure.
Strengths
- Native multimodal (text + vision)
- MoE architecture — modest active parameter count for inference
- Open weights with the broad Llama ecosystem
Weaknesses
- Newer architecture — fewer fine-tunes vs the Llama 3 line
- Community License restricts use at >700M MAU
Try it
| Where | Type | Notes |
|---|---|---|
| Hugging Face | weights | Gated — accept Community License |
| OpenRouter | hosted-api | API key required |
Used in solutions
Version history
- Llama 4 Maverick Apr 2025
- Llama 4 Scout Apr 2025 Current
Official sources
- Model card model
Change log
- — Added hardware requirements (Mac support, memory, NVIDIA). Source: estimate from 109B size at 4-bit.
- — Release date corrected to 2025-04-05, Meta's actual Llama 4 launch. The previous 2025-05-22 was the Hugging Face repo push timestamp. Still the newest Llama — there is no Llama 5; Meta's open-weight work moved to the Muse family (see Muse Glimmer 30B).
- — Initial entry. Sighted on Hugging Face (lastModified 2025-05-22).
Esc