Agent benchmarks

Quant-Arena reports that only 4 of 9 tested models made money on unseen market data

EinsiaAI released Quant-Arena, a benchmark for end-to-end quantitative research using A-share market data, reporting that only 4 of 9 frontier models made money. The test covered 3,042 stocks and a sealed 2024 evaluation year. KaiteeShiks said that agents wrote Python, built signals and developed strategies before testing on unseen data with trading costs included.

ryanCarerhq1 said that the equal-weight benchmark beat every model, while 88C888 reported a 17.1% return for equal weighting against 10.7% for the best model. Melisasmith_ highlighted a gap between stock ranking and returns: GLM-5.3 recorded a RankIC of 0.046 but lost 7.8%, while Claude Opus 5.5 had a RankIC of -0.002 and returned 4.4%.