Llama 4 Scout: Meta's 10M-Token Open-Weight Model
Key facts · verified Oct 29, 2026
Llama 4 Scout is a Meta model released on 5 Apr 2025. Pricing: Open weights, self-hosted. Its context window is 10M tokens. It ranks #262 on LMArena's text leaderboard (Elo 1321, read 3 Nov 2026). Source: ai.meta.com.
Llama 4 Scout is Meta's open-weight multimodal model, announced on April 5, 2025 with a 10M-token context window. It uses a mixture-of-experts design with 17B active and 109B total parameters across 16 experts. The weights are downloadable from llama.com and Hugging Face.
Key Features
What is Llama 4 Scout?
Llama 4 Scout is an open-weight multimodal model from Meta, announced 5 April 2025, with a 10 million token context window. Meta describes a mixture-of-experts architecture with 17 billion active parameters and 109 billion total across 16 experts, taking text and image input.
Licensing
The weights are downloadable from llama.com and Hugging Face. Meta's announcement does not name a specific licence; the terms are published on those download pages.
Ranking and Context Against Other Models
Context windows: Llama 4 Scout 10 million tokens, Claude Opus 5 1 million, GPT-5.6 1.05 million, Mistral Large 3 256k. On LMArena's text leaderboard Llama 4 Scout has 1321 Elo (rank 262), Claude Opus 5 1492 Elo (rank 14, high effort) and Mistral Large 3 1414 Elo (rank 146).
Key Takeaways
- Meta's open-weight model with a 10M token context window
- Meta's announcement does not name a licence; the terms are on llama.com and Hugging Face
- Mixture-of-experts: 17B active of 109B total parameters, 16 experts
- Text and image input
- LMArena: 1321 Elo (rank 262)
Official Resources
Full Specifications
Open weights, self-hosted | |
|---|---|
| Identity | |
| Developer | Meta |
| Released | Apr 2025 |
| Status | GA |
| Licence | Open weights (Llama 4 Community License Agreement) |
| Self-hostable | Yes |
| Cost | |
| Input price | Not published by Meta; Llama 4 Scout is an open-weights release and per-token prices are set by the host serving it. |
| Output price | Not published by Meta; Llama 4 Scout is an open-weights release and per-token prices are set by the host serving it. |
| Cached input | Not offered |
| Batch discount | 50% off input/output (AWS Bedrock Batch: $0.085 / $0.33 per 1M tokens) |
| Free tier | Not offered |
| Capacity | |
| Context window | 10M tokens |
| Max output | 8K tokens |
| Long-context surcharge | None (flat AWS Bedrock rate across the 10M window) |
| Capability | |
| Vision in | Yes |
| Audio in | No |
| Function calling | Yes |
| Structured output | Yes |
| Extended reasoning | No |
| Web search | Not offered |
| Code execution | Not offered |
| Access | |
| API | No |
| Chat app | Yes |
| Cloud marketplaces | AWS Bedrock, Azure AI Foundry, Google Cloud Vertex AI |
| Fine-tuning | Not offered |
| Measured quality | |
| LMArena Elo | 1321 (checked Nov 2026) |
| LMArena rank | Rank 262 |