Llama 4 Scout

Meta · released Apr 5, 2025

Open weightsTool useStructured outputVision

Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B. It supports native multimodal input...

Compare this model Source listing

Specification

Context window: 1.31M
Max output: 16K
Knowledge cutoff: Aug 31, 2024
Parameters: 108.6B
Licence: other
Serving providers: 4
Moderated: No
Uptime: 100.0%

Intelligence

10.3

36th percentile

Coding

8.2

Coding Index

Agentic

1.1

Agentic Index

Output speed

152 t/s

Median across providers

Latency

265ms

Time to first token

Cost per task

$0.01

Estimated

Benchmarks

Where the score comes from

The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.

Evaluation scores

Percentage correct · higher is better

MMLU-Pro
75.2%
GPQA Diamond
58.7%
IFBench
39.5%
AA-LCR (long context)
30.3%
LiveCodeBench
29.9%
SciCode
17.0%
τ²-bench (Telecom)
15.5%
AIME 2025
14.0%
Humanity's Last Exam
3.8%
Terminal-Bench Hard
1.5%

An evaluation missing from this list was not run for this model — it is not a zero.

View as table

Evaluation	Score
MMLU-Pro	75.2%
GPQA Diamond	58.7%
IFBench	39.5%
AA-LCR (long context)	30.3%
LiveCodeBench	29.9%
SciCode	17.0%
τ²-bench (Telecom)	15.5%
AIME 2025	14.0%
Humanity's Last Exam	3.8%
Terminal-Bench Hard	1.5%

Against its peers

Intelligence Index · this model highlighted, nearest peers in grey

GPT-4o (2024-11-20)OpenAI
11.1
GPT-4oOpenAI
11.1
Qwen3 VL 32B InstructQwen
11.0
GLM 4.6VZ.ai
10.9
Llama 4 ScoutMeta
10.3
Qwen3 VL 30B A3B InstructQwen
9.9
R1 Distill Llama 70BDeepSeek
9.8
GPT-4.1 NanoOpenAI
9.6

Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.

View as table

Model	Intelligence
GPT-4o (2024-11-20)	11.1
GPT-4o	11.1
Qwen3 VL 32B Instruct	11.0
GLM 4.6V	10.9
Llama 4 Scout	10.3
Qwen3 VL 30B A3B Instruct	9.9
R1 Distill Llama 70B	9.8
GPT-4.1 Nano	9.6

Percentile among all indexed models

Intelligence36th

Coding8th

Agentic3th

Mathematics17th

Terminal work15th

Arena Elo5th

Pricing

What it costs to run

List prices per million tokens, plus what one representative task works out to.

List price

Input / 1M tokens: $0.1
Output / 1M tokens: $0.3
Cached input / 1M: Not offered
Blended 3:1: $0.15

One task, estimated

$0.01

Input tokens: 50,000
Output tokens: 25,000
Profile: Standard

Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).

Serving providers

Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.

Arena

Head-to-head generation quality

Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.

Overall Elo: 923
Win rate: 39.3%
Strongest at: Data visualisation
Tournaments: 56

Elo by category

Dot position on a 750–925 scale · Elo has no meaningful zero

750925

Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.

View as table

Category	Elo	Win rate
Data visualisation	923	39.3%
Game development	814	27.4%
UI components	803	25.5%
Websites	773	22.7%

Llama 4 Scout

Specification

Where the score comes from

Evaluation scores

Against its peers

Percentile among all indexed models

What it costs to run

List price

One task, estimated

Serving providers

Head-to-head generation quality

Elo by category

More from Meta

Llama 4 Maverick

Llama Guard 4 12B

Llama 3.3 70B Instruct

Llama 3.2 1B Instruct

Llama 3.2 3B Instruct

Llama 3.1 70B Instruct