Llama 4 Scout

Apr 2025
Llama 4 Scout
Compare in
on
$0.11per task?
$0.10 / $0.30per 1M in / out?
89 tok/sfirst token in 0.8 s?
10Mcontext

Rankingin non-thinking mode?

Strong at

C
Coding#78 of 137Agents & work tasks#175 of 177Tool calling#210 of 251Reasoning#337 of 360Science#234 of 336Long context#226 of 330Knowledge#228 of 332Instructions#161 of 252Vision#108 of 132

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperGPT-5 Minihigh×2.8-52%$0.053Context: 10M → 400KBetterGPT OSS 20BhighOpen×2.6-91%$0.011Context: 10M → 131K+1OpenGPT OSS 120Bhigh×2.9-4%$0.11Context: 10M → 131K+2CheaperGPT-5 Nanohigh×2.6-8%$0.05 / $0.40Context: 10M → 400KCheaperQwen3 235B A22B Thinking 2507thinkingOpen×2.6-29%$0.08Context: 10M → 262K+3

Run it yourself

estimate
Fits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✗ 48 GB GPU✓ 80 GB GPU · compressed
LicenceLlama 4 licencecommercial use with conditions

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam3.8#346
AA Omniscience Index-52.2#232
CritPt (physics)0#184
AA Intelligence Index 4.3.26.5#306
AA Omniscience · Accuracy15.2#242
AA Omniscience · Hallucination Rate79.4#170
AA-LCR (long context) 1.127.7#228
GDP.pdf · AA all-pass0#137
MMLU-Pro · official boardlab74.3
GPQA Diamond58.7#237
AIME 2025 · AA archived14#132
Coding
Terminal-Bench 4.00#78
SciCode21.3#140
Terminal-Bench Hard (version unstated)1.5#203
Terminal-Bench 2.13.7#143
LiveCodeBench · AA archived29.9#126
Agents & Tool Use
AutomationBench-AA0.2#135
GDPval-AA (work tasks) 2.1-163.7#175
τ³-Banking · AA3.3#132
AA-Briefcase 1.1 · Elo0#133
τ²-bench · AA15.5#215
Writing & Language
IFBench (instructions)39.5#163
Vision & Documents
MMMU-Pro (vision)52.9#111

FAQ

How much does Llama 4 Scout cost?›

$0.10 per 1M input tokens and $0.30 per 1M output tokens (cheapest host: DeepInfra). A typical benchmark task costs about $0.11 at default effort.

Is Llama 4 Scout free to use?›

There is no free API; the cheapest paid price is $0.10 per 1M input tokens. The weights are open, so you can run it yourself.

How good is Llama 4 Scout, and what is it best at?›

#337 of 360 in Chat & Knowledge, #78 of 137 in Coding, #134 of 138 in Agents & Tool Use, #154 of 197 in Writing & Language, #108 of 132 in Vision & Documents.

Is Llama 4 Scout better than Ministral 3 8B 2512?›

On LiveCodeBench · AA archived, Llama 4 Scout scores 29.9 vs 30.3 — about the same. It is 912% more expensive per task.

How can I use Llama 4 Scout?›

Through the API of DeepInfra, Google, Novita.

What is the context window of Llama 4 Scout?›

10M tokens, with up to 16K output tokens.

When was Llama 4 Scout released? Is there a newer version?›

Released 2025-04-05. It is the current version in its line.

Is Llama 4 Scout open source?›

Yes, the weights are open (llama4 licence).

Can I run Llama 4 Scout locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

Can Llama 4 Scout read images?›

It accepts text and images as input.

What is Llama 4 Scout?›

Llama 4 Scout is an open-weights language model from Meta, released Apr 2025.

What are the best alternatives to Llama 4 Scout?›

GPT-5 Mini (cheaper), GPT OSS 20B (better), GPT OSS 120B (open weights).