Bug Hunt Bench ranks frontier coding models on 105 planted real-repo bugs
PawelHuryn · x · 2026-09-11
Pawel Huryn launched Bug Hunt Bench, a blind-graded leaderboard pitting frontier coding models against 105 real planted bugs in actual repos, with one prompt per repo and continuous updates.
- The board offers score-vs-cost and score-vs-time views: costs span roughly 200x (log axis required), while runtimes stay under 7x (linear axis).
- Features include ceiling-run filtering, best-per-lab views, superseded historical runs, and PNG export; the same model can appear multiple times across reasoning tiers and harnesses.
- The author also shared a view with all OpenAI models selected, responding to a question about the Terra xhigh tier (it exists, just hidden by default filters).
More from Models
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11
- User Hails Unconfirmed 'DeepSeek 4.1 Flash' as an Inflection Point in LLMs — himanshustwts · 2026-09-11
- Terminal Bench v4: GLM-5.3 Leads at 41.9%, Kimi-K3 Underwhelms at 12.6% — Ok_Warning2146 · 2026-09-11
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11