NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/Llama 4 Scout
Meta logo

Llama 4 Scout

by Meta · Released Apr 2025

Open SourceMultimodal1M Context
16.5
avg score
Rank #290
Compare
Better than 7% of all models
Context
1.3M tokens (~655 books)
Input $/1M
$0.10
Output $/1M
$0.30
Type
multimodal
License
Open Source
Benchmarks
23 tested
Data as of
About

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...

Tested on 23 benchmarks · BenchGecko score 16.5. Top scores: MATH level 5 (62.3%), Artificial Analysis · GPQA Diamond (58.7%), Artificial Analysis · MMMU Pro (52.9%).

Capabilities
coding
9.1
#191 globally
reasoning
0.3
#228 globally
math
23.3
#214 globally
knowledge
35.9
#222 globally
speed
18.3
#108 globally
general
21.9
#157 globally
Benchmark Scores
Compare All
Tested on 23 benchmarks · Ranked across 6 categories
Score Distribution (all 312 models)
0255075100
▲ You are here
codingCompare coding →
SWE-Bench Verified (Bash Only)

SWE-bench Verified solved using only bash commands, no specialized frameworks. Tests raw terminal-based problem solving.

9.1·
reasoningCompare reasoning →
ARC-AGI

Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.

0.5·
ARC-AGI-2

ARC-AGI 2, harder sequel to ARC. More complex abstract reasoning patterns that test generalization ability beyond training data.

0.1·
mathCompare math →
MATH level 5

Competition-level math from AMC, AIME, and olympiad problems. Level 5 is the hardest tier, requiring creative problem-solving.

62.3·
OTIS Mock AIME 2024-2025

Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.

7.7·
FrontierMath-2025-02-28-Private

Original research-level math problems created by professional mathematicians. Problems are unpublished and cannot be memorized.

0.0·
Quick compare:Alibaba logovs Qwen2.5 Coder 0.5B Instructvs Magistral Small 1.1Alibaba Qwen logovs Qwen-1_8B
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Model Family · Meta Llama 4
Apr 2025
22.1
$0.19/M in1.0Mctx29 benchmarks
Apr 2025
16.5-5.6
$0.10/M in(-0.09)1.3Mctx(+262K)23 benchmarks
Apr 2025
n/an/actx1 benchmark
See the full Llama 4 family →
Similar Models
Alibaba logo
Qwen2.5 Coder 0.5B Instruct
Alibaba
16.0n/a
Magistral Small 1.1
Unknown
17.0n/a
Alibaba Qwen logo
Qwen-1_8B
Alibaba Qwen
15.9n/a
Links
Info
MetaPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@Meta
Source Code
GitHubHugging Face
BenchGecko API
llama-4-scout
Specifications
  • Typemultimodal
  • Context1.3M tokens (~655 books)
  • ReleasedApr 2025
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.001
Available On
Meta logoMeta$0.10
Categories
codingreasoningmathknowledgespeedgeneral
Learn More
context-windowtransformeropen weightstokens
Share & Export
Tweet
Related Models
Alibaba logoQwen2.5 Coder 0.5B Instruct
16.0
Magistral Small 1.1
17.0
Alibaba Qwen logoQwen-1_8B
15.9
DeepSeek logoDeepSeek Coder 6.7B
17.1
Llama 4 Scout is an open-source multimodal AI model by Meta, released in April 2025. It has an average benchmark score of 16.5. Context window: 1M tokens.

Related Models

Qwen2.5 Coder 0.5B Instruct · AlibabaMagistral Small 1.1 · UnknownQwen-1_8B · Alibaba QwenDeepSeek Coder 6.7B · DeepSeekPhi-1.5 · Microsoft

Benchmarks

MATH level 5Artificial Analysis · GPQA DiamondArtificial Analysis · MMMU ProArtificial Analysis · IFBenchFiction.LiveBench

Related Pages

Meta · ProviderMeta · EconomyAll ModelsCompare ModelsPricingDevelopers · APIContext Window · Glossary

Key facts · as of 2026-10-05

  • Llama 4 Scout by Meta. BenchGecko score 16.5, rank 290 of 312 scored models (normalized average of public benchmark scores).
  • List price $0.10 input · $0.30 output per 1M tokens (as of 2026-10-05).
  • Sold by 3 providers (as of 2026-10-05): DeepInfra (fp8) $0.10 in / $0.30 out · Novita (bf16) $0.18 in / $0.59 out · Google $0.25 in / $0.70 out. Every provider
  • Gecko Tests: Who Are You D (Sometimes says it is OpenAI) · Tokenizer Tax B (43% more tokens outside English).

How to cite · data as of 2026-10-05

Llama 4 Scout · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-05. https://benchgecko.ai/model/llama-4-scout

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP