Skip to content
Longterm Wiki

Llama 4 Scout

Meta AI (FAIR)Open Weight

Llama 4 Scout was released April 5, 2025 as a 17B active parameter mixture-of-experts model (109B total, 16 experts). Featured a 10M token context window — the longest of any production model at launch. Natively multimodal (text + image). Scored 89.3% on MMLU. Beat Gemini 2.0 Flash and GPT-4o on multiple benchmarks while being deployable on a single H100 GPU.

Developer
Meta AI (FAIR)
Released
2025-04-05
Context Window
10M tokens

Benchmarks6

Knowledge

2 benchmarks
89.3%#11/37
70th percentile
74.3%#10/15
33th percentile

Reasoning

1 benchmark
57.2%#20/34
43th percentile

Coding

1 benchmark
32.8%#9/9
6th percentile

Multimodal

2 benchmarks
94.4%#1/2
50th percentile
69.4%#3/3
17th percentile
Percentile among tested models:Top 25%50-75%25-50%Bottom 25%

Llama Family5

ModelTierReleasedInput $/MTok
Llama 4 Maverick2025-04-05
Llama 3.32024-12-06
Llama 3.12024-07-23
Llama 32024-04-18
Llama 22023-07-18

Details

Model FamilyLlama
Generation4
Release Date2025-04-05
Parameters109B
Context Window10M tokens
Open WeightYes
Modalitytext, image

Capabilities3

tool-usevisionlong-context

Sources1

Tags

llamametaopen-weightmixture-of-expertslong-context