Llama 4 Scout — Specifications
| Developer | Meta |
|---|
| Type | Multimodal (MoE) |
|---|
| Modality | Text, Image → Text |
|---|
| Parameters | 109B total / 17B active (MoE) |
|---|
| Context window | 10M |
|---|
| Max output | — |
|---|
| License | Llama 4 Community (EU-restricted) |
|---|
| Open weights | ✅ Yes |
|---|
| Released | 2025 |
|---|
| Input price | $0.1 /1M |
|---|
| Output price | $0.3 /1M |
|---|
| API providers | Meta, Together, OpenRouter |
|---|
🖥️ Run it locally
| VRAM (4-bit) | ~65 GB |
|---|
| Minimum GPU | H100 80GB / Mac 128GB |
|---|
Official page →
Meta’s natively-multimodal open MoE — 109B total / 17B active (16 experts) with an industry-leading 10M-token context. Fits a single 80GB GPU at 4-bit. Llama 4 Community License (restricted in the EU).
