arxiv
GPT-5 mini + CodeStruct
1 benchmarks
SWE-bench Verified
#64 of 86
62.0%
benchmark.space canonical public ledger of AI capability