Bug Hunt Bench

A blind-graded coding-agent benchmark with 105 planted bugs across two production TypeScript repositories.

ModelsModels with a result.25
Top resultBest result on this benchmark.58.0%Claude Fable 5.2
Top-3 spreadPoints from first to third.13.0 pts
YearYear of release.2026

Result and price

23 of 25 results

Results

25 results
#ModelResultThe score from the source.
14Grok 4.729.0%
15Grok 4.627.0%
20Kimi K321.0%
21GLM-5.319.0%
22GPT-6 Luna18.3%