Llama 4 Scout | Epoch AI

ECI#152

130

Meta AI

Apr. 5, 2025

Open weights

Llama 4 Scout is an open-weights model from Meta AI, released April 5, 2025. It has an ECI score of 130 and ranks 152nd of 249 tracked models. At the time of its release, it ranked 25th of 121 tracked models.

Llama 4 Scout vs select alternatives

Domain	Benchmark	Llama 4 Scout Meta AI Apr. 5, 2025 Open weights	Llama 3.1-405B Meta AI Jul. 23, 2024 Open weights	GPT-4.1 nano OpenAI Apr. 14, 2025 Closed weights	Gemma 3 27B Google DeepMind Mar. 12, 2025 Open weights
	ECI scoreBest score 166	130	129	130	130
	Ranking	#152	#159	#153	#150
Conceptual reasoning	ARC-AGI-2Best score 95%	0%	n/a	0%	n/a
	ARC-AGI-1Best score 99%	1%	n/a	0%	n/a
Long context	ForecastBenchBest score 62.5	57.5	59.9	—	—
	Fiction.liveBenchBest score 97%	36%	n/a	25%	33%
Mathematics	MATH Level 5Best score 98%	62%	50%	70%	74%
	OTIS Mock AIME 2024-2025Best score 100%	8%	10%	29%	23%
Science	CritPtBest score 32%	0%	n/a	0%	0%
	SciCodeBest score 62%	17%	n/a	26%	21%
	GPQA DiamondBest score 96%	52%	51%	49%	48%

Model metadata

General
Organization	Meta AI
Release date	Apr. 5, 2025
Model accessibility	Open weights
Country	United States of America
Knowledge cutoff	2024-08-01
Provider notes	The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation

Pricing
Input price	$0.105 / 1M tokens
Output price	$0.465 / 1M tokens

Training
Parameters	1.1e11
Training compute	4.1e24 FLOP
Training hardware	Unknown
Training time	Unknown
Training dataset size	3.0e13 datapoints