An
claude-opus-4-7
Benchmark & Insights
Anthropic Claude API
Updated Oct 4, 2026 All models
Sample size
222 runs
in window
Accuracy
73.6%
reference match · 231d
Confidence
86%
over 232 runs
Window end
Oct 4, 2026
most recent run
Input price
$5.00/MTok
prompt tokens
Output price
$25.00/MTok
completion tokens
Model insights
- 01 Has 10 days missing, mostly in August and September, so its score covers slightly fewer hard days.
- 02 It was the best model over kimi-k2.5's spring stretch, but on the full period it leans low, calling 35 "MEDIUM_SAFE" days "low" and missing 11 of 23 "unsafe" days.
- 03 Weak in March (53%), when hard days were common.