An
claude-opus-5
Benchmark & Insights
Anthropic Claude Api
Updated Sep 12, 2026 All models
Sample size
211 runs
in window
Accuracy
91.1%
consensus match · 214d
Confidence
84%
over 214 runs
Window end
Sep 12, 2026
most recent run
Input price
$5.00/MTok
prompt tokens
Output price
$25.00/MTok
completion tokens
Model insights
- 01 The newest Opus scores below every older Opus at the same price.
- 02 It is the most optimistic Claude, with essentially every miss a "low" or "medium" call under the consensus, and it said "safe" on six "unsafe" days, four of them "high" risk.
- 03 Misses pile up in July and August, including a three-day run in mid-July, so it has drifted downward recently.