The LLM benchmark aggregator.

Claude Opus 4.5

Anthropic

Strongest in Tool use (#1 of 81), weakest in Safety (#203 of 337). Above par in 19 of 24 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5 and Claude Fable 5 and ahead of Claude Sonnet 4.6 and Grok 4.

§ 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.

Agents63.3−4.8#13/2681/5

Tool use74leads#1/811/1

Professional56.9−4.7#19/1681/1

Medical58.8−8#15/1401/1

Finance57.2−5.7#27/1511/1

Legal55.8−11.3#35/1511/1

Human preference64.1−3.4#31/3421/1

Knowledge56.3−7.2#34/1381/2

Academic knowledge57.5−6.4#31/1231/1

Reasoning53.6−13.3#54/1784/4

Mathematics56.1−9.6#37/1402/2

Science56.8−6.7#45/1221/1

Reasoning50.3−18.8#60/1403/3

Core abilities52−15.4#71/2042/3

Language53.8−17.7#23/571/1

Data analysis51.7−14.4#31/571/1

Instruction following40.5−33.2#45/571/1

General intelligence57−13.1#46/2042/3

Coding46.4−22.7#106/1652/5

Code generation54.7−15.3#29/771/2

Agentic coding43.6−25.1#117/1572/4

Safety49.3−12.2#203/3371/3

Factual grounding47.7−22.9#68/1011/1

§ 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 Claude Opus 4.5, left for the other.

Qwen3.6 27Bwon 15 · lost 0

Mistral Medium 3.5won 16 · lost 0

Mistral Large 3won 19 · lost 0

GPT-4.1 Nanowon 18 · lost 1

Command Awon 18 · lost 1

GPT-4o Miniwon 17 · lost 1

Llama 4 Scout Instructwon 17 · lost 1

Nemotron 3 Ultrawon 15 · lost 1

Mistral Smallwon 15 · lost 1

GPT OSS 20Bwon 14 · lost 1

Grok 4.3won 20 · lost 2

Gemini 3.5 Flash Litewon 19 · lost 2

Llama 4 Maverick Instructwon 18 · lost 2

DeepSeek V3won 18 · lost 2

Claude Haiku 4.5won 18 · lost 2

GPT OSS 120Bwon 17 · lost 2

GLM-4.6won 17 · lost 2

GPT-4owon 15 · lost 2

Gemini 2.0 Flashwon 15 · lost 2

Gemini 2.5 Flash Litewon 15 · lost 2

MiniMax M2.5won 14 · lost 2

Grok 3won 14 · lost 2

MiniMax M3won 20 · lost 3

GPT-5.4 Miniwon 19 · lost 3

Qwen3.6 Pluswon 18 · lost 3

Kimi K2won 17 · lost 3

DeepSeek V3.2won 17 · lost 3

GPT-4.1 Miniwon 16 · lost 3

Kimi K2.5won 16 · lost 3

GPT-5 Nanowon 15 · lost 3

Grok 4 Fastwon 15 · lost 3

Grok 4.1 Fastwon 15 · lost 3

GPT-5.4 Nanowon 19 · lost 4

Inklingwon 19 · lost 4

GLM-5.2won 19 · lost 4

Mimo V2.5 Prowon 14 · lost 3

Grok 3 Miniwon 13 · lost 3

Claude 3.7 Sonnetwon 13 · lost 3

Gemini 2.5 Flashwon 17 · lost 4

Kimi K2 Thinkingwon 12 · lost 3

O1won 12 · lost 3

Inkling Smallwon 12 · lost 3

GLM-4.7won 12 · lost 3

GPT-6 Lunawon 15 · lost 4

Gemini 3.1 Flash Litewon 14 · lost 4

O3 Miniwon 13 · lost 4

GPT-5 Miniwon 15 · lost 5

GPT-4.1won 15 · lost 5

Claude Sonnet 4.5won 15 · lost 5

DeepSeek R1won 14 · lost 5

Claude Opus 4won 14 · lost 5

Claude 3.5 Sonnetwon 11 · lost 4

MiniMax M2.7won 11 · lost 4

Qwen3 Maxwon 11 · lost 4

O4 Miniwon 15 · lost 6

Gemma 4 31Bwon 12 · lost 5

Claude Sonnet 4won 14 · lost 6

Qwen3 235B A22B Instructwon 14 · lost 6

O3won 14 · lost 6

Kimi K2.6won 16 · lost 7

Qwen3.8 27Bwon 15 · lost 7

DeepSeek V4 Flashwon 15 · lost 7

GPT-5.6 Lunawon 15 · lost 8

GLM-5.3 Flashwon 14 · lost 8

Qwen3.7 Maxwon 14 · lost 8

GLM-5.1won 10 · lost 6

GLM-5won 11 · lost 7

Qwen3.5 Flashwon 9 · lost 6

GPT-5won 11 · lost 8

Gemini 3 Flashwon 9 · lost 7

Gemini 2.5 Prowon 11 · lost 9

Grok 4won 10 · lost 9

Claude Sonnet 4.6won 12 · lost 11

Gemini 3.6 Flashwon 11 · lost 11

GPT-5.2won 12 · lost 12

Claude Sonnet 5won 11 · lost 12

DeepSeek V4 Prowon 11 · lost 12

GPT-5.6 Terrawon 11 · lost 12

GLM-5.3won 10 · lost 11

GPT-5.1won 8 · lost 9

Grok 4.7won 8 · lost 11

Grok 4.5won 9 · lost 13

Gemini 3.5 Flashwon 9 · lost 13

Muse Spark 1.1won 9 · lost 13

DeepSeek V4.1 Flashwon 8 · lost 12

GPT-6 Solwon 7 · lost 13

Gemini 3 Prowon 7 · lost 13

Gemini 3.1 Prowon 8 · lost 15

Grok 4.6won 7 · lost 15

GPT-5.4won 7 · lost 16

Gemini 3.7 Flashwon 6 · lost 15

Qwen3.8 Maxwon 6 · lost 16

Gemini 3.8 Flashwon 6 · lost 16

Muse Spark 1.2won 5 · lost 15

Claude Opus 4.6won 5 · lost 17

GPT-6 Astrawon 4 · lost 16

GPT-5.6 Solwon 4 · lost 19

Claude Opus 4.8won 3 · lost 19

GPT-5.5won 3 · lost 20

Claude Opus 4.7won 3 · lost 20

Kimi K3won 2 · lost 19

Claude Opus 5.5won 1 · lost 18

Muse Spark 1.3won 0 · lost 17

Claude Fable 5.1won 0 · lost 21

Claude Fable 5won 0 · lost 21

Claude Opus 5won 0 · lost 22

§ 3 · Sources

Where the numbers come from

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

LMArena Text ↗1 measureread 2026-09-26

LMArena Text 1474

ARC-AGI-2 ↗1 measureread 2026-09-26

ARC-AGI-2 37.6%

LiveBench ↗8 measuresread 2026-09-26

LiveBench 72.6LiveBench · Reasoning 80.1LiveBench · Coding 79.7LiveBench · Agentic Coding 39.7LiveBench · Mathematics 90.4LiveBench · Data Analysis 74.4LiveBench · Language 81.3LiveBench · Instruction Following 62.5

BFCL ↗1 measureread 2026-09-26

BFCL v4 77.47%

Kagi LLM Benchmark ↗1 measureread 2026-09-26

Kagi LLM Benchmark 80.2%

SimpleBench ↗1 measureread 2026-09-26

SimpleBench 62%

Vals.ai ↗13 measuresread 2026-09-26

Vals · CaseLaw 62.59%Vals · LegalBench 84.6%Vals · CorpFin 65.07%Vals · TaxEval 74.86%Vals · MortgageTax 67.69%Vals · MedQA 95.88%Vals · MedCode 49.16%Vals · MedScribe 85.32%Vals · SWE-bench Verified 76.4% (Mini-SWE-agent)Vals · Vibe Code Bench 20.63% (OpenHands)Vals · GPQA Diamond 85.86%Vals · MMLU Pro 87.26%Vals · AIME 95.42%

Vectara Hallucination Leaderboard ↗1 measureread 2026-09-26

Vectara · Factual consistency 89.1%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=claude-opus-4-5)](https://publicai.io/model-index/m/claude-opus-4-5)