An
claude-fable-5
Benchmark & Insights
Anthropic Claude API
Updated Oct 4, 2026 All models
Sample size
65 runs
in window
Accuracy
80.7%
reference match · 114d
Confidence
88%
over 114 runs
Window end
Oct 4, 2026
most recent run
Input price
$10.00/MTok
prompt tokens
Output price
$50.00/MTok
completion tokens
Model insights
- 01 Ran only since 2026-08-01, but on those days it still beat most peers, though claude-sonnet-4-6 was clearly better.
- 02 Zero false alarms, and it caught 5 of 7 "unsafe" days; both misses (2026-09-25 and 2026-09-26) were hard days.
- 03 At $10/$50 it is the most expensive model and does not beat cheaper Sonnets.
Notes
Anthropic released a new class of models Fable 5
Recent forecasts