VerdictPal
Join

Models / Meta

Llama 4 Scout

Llama 4 · Released 2026-04-05

Meta's long-context Llama 4 MoE. 10M token window, 17B active / 109B total params. Open weights for codebase-scale RAG and document mining.

Self-hosted route when you need entire repos or document libraries in one pass — not for lowest-latency chat.

#47 of 51 on AA Index · current snapshot
10 AA Index · 18 benchmark rows · 6 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite41
Context10M
Input / 1M$0.08
Output / 1M$0.30
Knowledge cutoff2024-08-01
Output speed76 t/s
TTFT (default API)0.56s
Statussolid
KnowledgeReasoningMathCodingAgenticLong contextTool use
Knowledge
85
Reasoning
59
Math
84
Coding
30
Agentic
59
Long context
57
Tool use
74

64average across 7 tested families

Benchmark placements

Where Llama 4 Scout places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
LongBench v2Long context#2/ 1957LongBench v2 leaderboard2026-04-01
MATHMath#10/ 2084Artificial Analysis2026-07-21
LiveCodeBenchCoding#11/ 1330Artificial Analysis2026-07-21
MMLU-ProKnowledge#15/ 2175Artificial Analysis2026-07-21
AIMEMath#17/ 1914Artificial Analysis2026-07-21
BFCLTool use#17/ 2074Berkeley Function Calling Leaderboard V42026-04-15
LiveBenchReasoning#19/ 2159LiveBench leaderboard2026-04-15
MMLUKnowledge#21/ 2285Meta Llama 4 Scout model card2026-04-05
AA output speedPerformance#35/ 69Artificial Analysis2026-07-21
Terminal-BenchAgentic#40/ 6959Terminal-Bench 2.0 leaderboard2026-06-09
AA time to first tokenPerformance#56/ 69Artificial Analysis2026-07-21
IFBenchReasoning#56/ 5740Artificial Analysis2026-07-21
AA-LCRLong context#61/ 6426Artificial Analysis2026-07-21
Artificial Analysis Intelligence IndexReasoning#63/ 6710Artificial Analysis2026-07-21
Humanity's Last ExamReasoning#63/ 654Artificial Analysis2026-07-21
SciCodeCoding#63/ 6517Artificial Analysis2026-07-21
τ³-Bench BankingAgentic#64/ 643Artificial Analysis2026-07-21
GPQA DiamondReasoning#65/ 6959Artificial Analysis2026-07-21
Family context

The three highest-scoring carded models in each capability family. Where Llama 4 Scout shows up, it's highlighted.

KnowledgeLlama 4 Scout · 85
  1. 1
    GPT 5.5
    91
  2. 2
    Claude Opus 4.8
    91
  3. 3
    Claude Sonnet 4.6
    91
ReasoningLlama 4 Scout · 59
  1. 1
    Claude Mythos Preview
    95
  2. 2
    GPT-5.6 Sol
    94
  3. 3
    Kimi K3
    94
MathLlama 4 Scout · 84
  1. 1
    OpenAI o3
    99
  2. 2
    OpenAI o1
    97
  3. 3
    Gemini 3 Pro
    96
CodingLlama 4 Scout · 30
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticLlama 4 Scout · 59
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextLlama 4 Scout · 57
  1. 1
    Kimi K3
    75
  2. 2
    GPT-5.1
    75
  3. 3
    GPT-5.6 Sol
    74
Tool useLlama 4 Scout · 74
  1. 1
    Claude Sonnet 4.6
    89
  2. 2
    GPT 5.5
    87
  3. 3
    Claude Opus 4.8
    87
Newest receipts

Benchmark rows added to the public ledger for Llama 4 Scout in the last 120 days. Older rows live in the full table below.

18 recent
BenchmarkFamilyScoreSourceDays ago
AA-LCRLong context25.8%Artificial Analysis2
AA output speedPerformance76 t/sArtificial Analysis2
AA time to first tokenPerformance0.56sArtificial Analysis2
AIMEMath14%Artificial Analysis2
Artificial Analysis Intelligence IndexReasoning10%Artificial Analysis2
GPQA DiamondReasoning58.7%Artificial Analysis2
Humanity's Last ExamReasoning4.3%Artificial Analysis2
IFBenchReasoning39.52%Artificial Analysis2
LiveCodeBenchCoding29.9%Artificial Analysis2
MATHMath84.4%Artificial Analysis2
MMLU-ProKnowledge75.2%Artificial Analysis2
SciCodeCoding17%Artificial Analysis2
τ³-Bench BankingAgentic3.3%Artificial Analysis2
Terminal-BenchAgentic58.6%Terminal-Bench 2.0 leaderboard44
BFCLTool use74.1%Berkeley Function Calling Leaderboard V499
LiveBenchReasoning58.9%LiveBench leaderboard99
MMLUKnowledge85.2%Meta Llama 4 Scout model card109
LongBench v2Long context57.2%LongBench v2 leaderboard113
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Knowledge85
Reasoning59
Math84
Coding30
Agentic59
Long context57
Tool use74
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMiniMaxxAIMetaZhipuXiaomiMeituanCohere

Full benchmark ledger

Every catalog benchmark for Llama 4 Scout. Scores link to the original source; gaps mean no public row exists yet.

18 sourced rows

18 of 32 catalog benchmarks have a sourced row for Llama 4 Scout.

Knowledge

Reasoning

Math

Coding

Agentic

Long context

Tool use

Performance

Sources