Atlas

Benchmarks

← All benchmarks

SRE-Bench - OpenAI release pass@1

Code

Single-attempt software reverse-engineering success in OpenAI's SRE-Bench evaluation. Exact source run count is not restated; distinct from Vals's separately specified SRE task cohort.

Top models (higher is better)

ModelScore
GPT-6 Astra88.0
Loading Atlas data…