Beta
Classificação/Claude Opus 4
Anthropic

Claude Opus 4

por Anthropic · Lançado em 2025-05-22

40.1
pontuação média
$15.00/1M
Preço de entrada
$75.00/1M
Preço de saída
200K tokens (~100 books)
Janela de contexto
multimodal
Tipo

Tested on 18 benchmarks with 40.1% average. Top scores: MATH level 5 (85.0%), Aider polyglot (72.0%), GPQA diamond (68.3%).

Pontuações de benchmark

BenchmarkCategoriaPontuaçãoBar
MATH level 5math85.0
Aider polyglotcoding72.0
GPQA diamondknowledge68.3
SWE-Bench Verified (Bash Only)coding67.6
OTIS Mock AIME 2024-2025math64.4
Fiction.LiveBenchknowledge61.1
SimpleBenchreasoning50.6
GeoBenchknowledge49.0
DeepResearch Benchknowledge49.0
WeirdMLcoding43.4
Cybenchcoding38.0
ARC-AGIreasoning35.7
ARC-AGI-2reasoning8.6
VPCTknowledge7.0
GSO-Benchcoding6.9
HLEknowledge6.2
FrontierMath-2025-02-28-Privatemath4.5
FrontierMath-Tier-4-2025-07-01-Privatemath4.2

Modelos similares

Mistral AIMistral Medium 3
Mistral AI
40.0
Google DeepMindGemma 2 9B
Google DeepMind
40.0
AlibabaQwen2.5-Coder (1.5B)
Alibaba
40.3
GoogleGemini 1.5 Flash (Sep 2024)
Google
40.4

Anthropic Claude Opus Cronologia

Claude 3 Opus
33.7
N/AN/Actx8 benchmarks
Claude Opus 4
40.1+6.4
$15.00/M in200Kctx18 benchmarks
Claude Opus 4.1
38.5-1.6
$15.00/M in200Kctx13 benchmarks
Claude Opus 4.5
46.5+8.0
$5.00/M in(-10)200Kctx18 benchmarks
Claude Opus 4.6
55.5+9.0
$5.00/M in1.0Mctx(+800K)14 benchmarks