An
claude-sonnet-5-5
Benchmark & Insights
Anthropic Claude API
Updated Sep 30, 2026 All models
Sample size
229 runs
in window
Accuracy
94.3%
consensus match · 229d
Confidence
83%
over 229 runs
Window end
Sep 30, 2026
most recent run
Input price
$2.00/MTok
prompt tokens
Output price
$10.00/MTok
completion tokens
Model insights
- 01 Ties for the best full-season score and is the most even-handed Claude, with misses split between too high and too low.
- 02 It called only 3 of 18 "unsafe" days "safe", fewer than any Opus except opus-4-6.
- 03 Matches the best Opus at well under half the price.
Recent forecasts