Ctrl K

Benchmarks

Who leads each test, what it measures and what the results cost. Every value is an independent measurement at its exact tested setting; developer-reported launch scores stay on model pages.

Source: Artificial AnalysisIntelligence Index v4.3.227 tests + composite Index83 measured modelsCaptured 22 Sep – 17 Oct 2026

Leaderboard Overall

10 evaluations of reasoning, knowledge and agentic work, combined into one score. Methodology (opens in a new tab)

Measured 83 of 83 modelsMedian 25.3Top 5 within 4.9 points

Index points Higher is better
Claude Opus 5.5

Max, Default Fallback

57.6 points
Source for Claude Opus 5.5 (opens in a new tab)Captured

Leads Claude Sonnet 5.5 by 1.6 pointsCompare with Claude Sonnet 5.5

All benchmarks

Who leads each test and where Claude Opus 5.5 stands. Select one to open its leaderboard.

BenchmarkLeaderClaude Opus 5.5Top 5 within
Overall
Claude Opus 5.557.6
57.6#1 of 83
4.9 points
Reasoning
Kimi-K388.7%
84.7%#5 of 83
4.0 pts
Reasoning
Muse Wonder Full32.9%
31.7%#2 tied of 83
1.4 pts
Reasoning
Claude Opus 5.561.4%
61.4%#1 of 83
6.7 pts
Knowledge
Claude Opus 5.546.4
46.4#1 of 83
4.9 pts
Knowledge
GPT-6 Astra96.1%
Not measured
2.5 pts
Coding
Claude Opus 5.566.9%
66.9%#1 of 83
5.9 pts
Agents & tools
Gemini 4 Argon77.5%
69.5%#4 of 83
8.6 pts

Intelligence trade-offs

Intelligence Index against weighted average task cost, using the same tested setting on both axes.

Pareto frontier · shown models

Index axis starts at 35

Not plotted · Muse Wonder Full. Cost not reported for the selected configurations.

Claude Opus 5.5

Max, Default Fallback

Intelligence Index57.6
Cost / task$5.98
Source for Claude Opus 5.5 (opens in a new tab)Captured

Cost, time and tokens

Weighted average cost to run one Intelligence Index task.

USD per task Lower is better
  • Answer
  • Reasoning
  • Cache Write
  • Cache Hit
  • Input
Range across measured settings

Not reported · Muse Wonder Full

Claude Opus 5.5

Max, Default Fallback

$5.98
Source for Claude Opus 5.5 (opens in a new tab)Captured
  • Answer $0.70
  • Reasoning $1.68
  • Cache Write $1.07
  • Cache Hit $2.42
  • Input $0.10

Across 5 settings $0.55 Low · fallback – $5.98 Max · fallback

Reasoning trade-offs

How scores trade against cost across reasoning settings for Claude Opus 5.5 and comparison models. Cost per task reflects each benchmark’s own workload, so compare costs within a benchmark rather than across benchmarks.

Claude Opus 5.5

Intelligence Indexv4.3.2

Source (opens in a new tab)
Cost per task (USD) · log scale
Claude Opus 5.5Max
Score
57.6
Cost / task
$5.98

Terminal-Bench v4.0

Source (opens in a new tab)
Cost per task (USD)
Claude Opus 5.5Max
Score
59.6%
Cost / task
$13.11

Design Arena