Back to models

Claude Opus 4

ECI#90
143
Anthropic
May 22, 2025
Closed weights

Claude Opus 4 is a closed-weights model from Anthropic, released May 22, 2025. It has an ECI score of 143 and ranks 90th of 249 tracked models. At the time of its release, it ranked 4th of 139 tracked models. At $15/Mtok for input tokens and $75/Mtok for output tokens, it is more expensive per token than similarly-capable models.

Input price$15 / 1M tokens
Output price$75 / 1M tokens
ParametersUnknown
Training computeUnknown

Claude Opus 4 vs select alternatives

DomainBenchmark
Claude Opus 4
Anthropic
May 22, 2025
Closed weights
Anthropic
Feb. 24, 2025
Closed weights
OpenAI
Apr. 16, 2025
Closed weights
Google DeepMind
Mar. 31, 2025
Closed weights
ECI scoreBest score 166
143141147144
Ranking#90#103#60#82
Conceptual reasoning
ARC-AGI-2Best score 95.0%
8.6%
0.9%
6.5%
n/a
ARC-AGI-1Best score 98.5%
35.7%
28.6%
60.8%
33.0%
Mathematics
MATH Level 5Best score 98.1%
85.0%
91.2%
97.8%
95.6%
OTIS Mock AIME 2024-2025Best score 100.0%
64.4%
57.8%
84.4%
n/a
Science
Humanity's Last ExamBest score 46.5%
10.7%
8.0%
20.3%
18.2%
GPQA DiamondBest score 95.8%
76.3%
78.5%
81.8%
83.8%
Software engineering
METR Time HorizonsBest score 1044.8
100.460.4119.7
SWE-bench VerifiedBest score 83.5%
70.7%
61.0%
62.3%
n/a
Aider PolyglotBest score 88.0%
72.0%
64.9%
81.3%
72.9%

Model metadata