The LLM benchmark aggregator.

Llama 4 Scout Instruct

Meta · 109B · open weights

Strongest in Factual grounding (#37 of 101), weakest in Secure code (#254 of 274). Above par in 4 of 25 scopes. Among the models it meets almost everywhere, it finishes behind Claude Sonnet 5 and Inkling and ahead of Command A and Jamba 1.6 Large.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Knowledge37.9−25.6#123/1381/2

Academic knowledge35.5−28.4#108/1231/1

Core abilities44.6−22.8#150/2041/3

General intelligence42.2−27.9#155/2041/3

Reasoning36.7−30.2#164/1782/4

Reasoning42.7−26.4#102/1401/3

Science29.7−33.8#112/1221/1

Mathematics37.4−28.3#117/1401/2

Professional36.3−25.3#166/1681/1

Legal45.3−21.8#101/1511/1

Finance38.8−24.1#134/1511/1

Medical26.1−40.7#140/1401/1

Human preference49.5−18#200/3421/1

Safety48.9−12.6#210/3373/3

Factual grounding55.7−14.9#37/1011/1

Safe-prompt compliance53.3−7.1#44/821/1

Toxicity avoidance56.1−2.9#72/2721/1

Jailbreak resistance56.8−8.1#84/2721/1

Fairness46.1−23.9#171/3002/2

Harm refusal48.8−12.8#172/3002/2

Secure code29.3−37.2#254/2741/1

Agents40.2−27.9#237/2682/5

Tool use43−31#52/811/1

Knowledge work34.7−38.9#161/1781/2

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Llama 4 Scout Instruct, left for the other.

Jamba 1.6 Miniwon 12 · lost 3

Command R Pluswon 14 · lost 4

Gemini 1.5 Flashwon 13 · lost 4

Jamba 1.6 Largewon 11 · lost 4

Command Awon 12 · lost 10

Mistral Smallwon 10 · lost 11

GPT-4.1 Nanowon 11 · lost 13

GPT-4o Miniwon 10 · lost 12

Ling 3.0 Flashwon 6 · lost 10

Mistral Largewon 7 · lost 12

Gemini 1.5 Prowon 7 · lost 12

Grok 4.3won 8 · lost 14

Llama 3.3 Instruct Turbo (70B)won 5 · lost 10

Mistral Medium 3.5won 5 · lost 10

Claude 3.5 Haikuwon 6 · lost 12

DeepSeek R1won 7 · lost 16

MiniMax M3won 7 · lost 16

DeepSeek V3won 7 · lost 17

Grok 4.1 Fastwon 6 · lost 15

GPT OSS 20Bwon 6 · lost 16

GLM-4.6won 6 · lost 17

Kimi K2.5won 6 · lost 17

Kimi K2.6won 6 · lost 17

Qwen3.6 Pluswon 4 · lost 12

Gemini 3.5 Flash Litewon 4 · lost 12

Gemma 4 31Bwon 5 · lost 15

GPT OSS 120Bwon 6 · lost 18

O3 Miniwon 5 · lost 16

Grok 3won 5 · lost 16

GPT-5.4 Miniwon 4 · lost 13

DeepSeek V4 Flashwon 5 · lost 17

GLM-5.3 Flashwon 5 · lost 17

Mistral Large 3won 4 · lost 14

GLM-5.2won 5 · lost 18

GLM-4.5won 4 · lost 15

GPT-4owon 4 · lost 16

Qwen3.7 Maxwon 4 · lost 18

Gemini 3.5 Flashwon 4 · lost 18

Gemini 3.6 Flashwon 4 · lost 18

Grok 4won 4 · lost 19

DeepSeek V4 Prowon 4 · lost 19

Kimi K2 Thinkingwon 3 · lost 15

GPT-5.4 Nanowon 3 · lost 15

Gemini 2.5 Flashwon 4 · lost 20

Kimi K2won 4 · lost 20

Qwen3 235B A22B Instructwon 4 · lost 20

DeepSeek V3.2won 4 · lost 20

Llama 4 Maverick Instructwon 4 · lost 21

GPT-6 Lunawon 3 · lost 16

Gemini 2.0 Flashwon 3 · lost 18

Mimo V2.5 Prowon 3 · lost 18

GPT-5.1won 3 · lost 19

Gemini 3.8 Flashwon 3 · lost 19

GPT-5.5won 3 · lost 20

GPT-5.6 Solwon 3 · lost 20

Grok 4 Fastwon 2 · lost 14

GPT-5 Nanowon 3 · lost 21

GPT-4.1won 3 · lost 21

Gemini 2.5 Prowon 3 · lost 21

Claude Haiku 4.5won 3 · lost 22

Gemini 3.1 Flash Litewon 2 · lost 15

Mimo V2.6 Prowon 2 · lost 15

GPT-6 Solwon 2 · lost 18

GLM-5.1won 2 · lost 18

Grok 4.6won 2 · lost 20

Grok 4.5won 2 · lost 20

Qwen3.8 Maxwon 2 · lost 20

Claude Opus 5won 2 · lost 20

GPT-5.2won 2 · lost 21

GPT-5.6 Lunawon 2 · lost 21

GPT-5.6 Terrawon 2 · lost 21

GPT-5.4won 2 · lost 21

GPT-4.1 Miniwon 2 · lost 22

O4 Miniwon 2 · lost 22

GPT-5won 2 · lost 22

GPT-5 Miniwon 2 · lost 23

GPT-6 Astrawon 1 · lost 14

GLM-5won 1 · lost 15

Claude 3.5 Sonnetwon 1 · lost 17

Claude Opus 4.5won 1 · lost 17

Gemini 3.1 Prowon 1 · lost 17

O1won 1 · lost 18

Claude Opus 4.6won 1 · lost 18

Claude Opus 5.5won 1 · lost 18

Claude Sonnet 4.6won 1 · lost 20

Gemini 3 Prowon 1 · lost 20

Claude Opus 4won 1 · lost 21

Muse Spark 1.1won 1 · lost 21

O3won 1 · lost 22

Claude Opus 4.7won 1 · lost 22

Claude Sonnet 4won 1 · lost 23

Claude Sonnet 4.5won 1 · lost 24

DeepSeek V4.1 Flashwon 0 · lost 15

Muse Spark 1.2won 0 · lost 15

GLM-5.3won 0 · lost 16

Gemini 3.7 Flashwon 0 · lost 16

Kimi K3won 0 · lost 16

Claude Fable 5.1won 0 · lost 16

Claude Fable 5won 0 · lost 16

Qwen3.8 27Bwon 0 · lost 17

MiniMax M2.7won 0 · lost 19

Grok 4.7won 0 · lost 19

Gemini 2.5 Flash Litewon 0 · lost 20

Claude 3.7 Sonnetwon 0 · lost 20

Claude Opus 4.8won 0 · lost 22

Inklingwon 0 · lost 23

Claude Sonnet 5won 0 · lost 23

§ 3 · Sources

Where the numbers come from

9 publications, 27 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26

LMArena Text 1322

GDPval-AA ↗1 measureread 2026-09-26

GDPval-AA -178

ARC-AGI-2 ↗1 measureread 2026-09-26

ARC-AGI-2 0%

BFCL ↗1 measureread 2026-09-26

BFCL v4 28.13%

Kagi LLM Benchmark ↗1 measureread 2026-09-26

Kagi LLM Benchmark 36.9%

Vals.ai ↗10 measuresread 2026-09-26

Vals · LegalBench 72.04%Vals · CorpFin 46.78%Vals · TaxEval 55.19%Vals · MortgageTax 57.75%Vals · MedQA 50.9%Vals · MedCode 23.31%Vals · MedScribe 50.59%Vals · GPQA Diamond 46.97%Vals · MMLU Pro 69.63%Vals · AIME 18.96%

HELM Safety ↗5 measuresread 2026-09-26

HELM Safety · HarmBench 60%HELM Safety · SimpleSafetyTests 97%HELM Safety · Anthropic Red Team 96.5%HELM Safety · BBQ 87.5%HELM Safety · XSTest 95.6%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26

Enkrypt · Jailbreak risk 7%Enkrypt · Harmful content risk 31.7%Enkrypt · CBRN risk 12.2%Enkrypt · Toxicity risk 2%Enkrypt · Bias risk 86.1%Enkrypt · Insecure code risk 75.6%

Vectara Hallucination Leaderboard ↗1 measureread 2026-09-26

Vectara · Factual consistency 92.3%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=llama-4-scout-instruct)](https://publicai.io/model-index/m/llama-4-scout-instruct)