Llama 4 Scout
Meta·
LLMsopen-weightlocalWorkstation
Overview
109B MoE (17B active) multimodal model with 10M token context window. Outperforms Gemma 3, Gemini 2.0 Flash, and Mistral 3.1 Small.
Capabilities and innovations
109B total / 17B active MoE10M context windowNatively multimodal16 experts, 1 activeOpen weights (Llama license)Early Fusion Architecture10M Token Context (Longest Open-Weight)Interleaved Attention / MoE Blocks
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 57.2% | unsourced |
| MMLU | 85.8% | unsourced |
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Verified: not reported by the vendor
- SWE-bench Pro: benchmark did not exist at release
- MMLU-Pro: not reported by the vendor
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
Architecture and hardware
- Parameters
- 109B total, 17B active per token (MoE)
- Estimated VRAM at Q4
- ~63 GB, Workstation class
Links
More from Meta
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.