Meta
Meta

Llama 4 Scout

2025-04-05

Llama 4 Scout is Meta's efficient multimodal language model with 16 experts activating 17 billion parameters out of 109B total. It supports native multimodal input (text and image) across 12 languages with a 10-million-token context window — one of the longest available — and uses early fusion for seamless modality integration. Designed for high efficiency and local or commercial deployment, it is instruction-tuned for multilingual chat, captioning, and image understanding tasks, released under the Llama 4 Community License.

API|Vision|Open ModelLlama
Knowledge Cutoff
2024-08-31
Input → Output Format
Context Memory
328KIN16KOUT
Cost/1M Words
$0.08IN$0.3OUT
Calculate Cost

AI Performance Evaluation

Arena Overall Score
1322
±5
As of 2026-04-23
Overall Rank
No.191
30,319 Votes
Arena by Ability
Hard Prompts
1329±6No.189
Expert Knowledge
1308±16No.185
Instruction Following
1299±7No.199
Conversation Memory
1320±9No.187
Creative
1290±10No.188
Coding
1361±9No.189
Math
1309±13No.177
Arena by Occupation
Creative Writing
1306±8No.186
Social Sciences
1336±9No.192
Media
1290±9No.180
Business
1319±9No.186
Healthcare
1343±15No.180
Legal
1344±14No.173
Software
1350±7No.193
Mathematics
1313±14No.177
Overall
AA Intelligence Index
14%↓25%
ForecastBench
54%↓5%
Reasoning & Math
AA Math Index
14%↓59%
GPQA Diamond
59%↓22%
HLE
4.3%↓13%
MMLU-Pro
75%↓7%
AIME 2025
14%↓60%
MATH-500
84%↓9%
Coding
AA Coding Index
6.7%↓27%
LiveCodeBench
30%↓36%
TAU2
16%↓58%
TerminalBench
1.5%↓30%
SciCode
17%↓24%
Language & Instructions
IFBench
40%↓17%
AA-LCR
26%↓36%
Hallucination (HHEM)
7.7%↓2%
Factual (HHEM)
92%↑2%
Output Speed
Standard Mode
136tok/s↑54
First Output 0.55s