One other thing @millions-crayon-32880, it seems like we had an eval failure for minimax-m2.7. I guess neither of us noticed this when we checked the PR. I’ve opened a PR to delete it from the openhands-index-results for now. It seems that inference ran OK, so it’s not clear if we’d need to run inference again, but if it’d be easier to run both inference and eval maybe we should do that. PR to remove the result for now so we’re not mis-reporting: https://github.com/OpenHands/openhands-index-results/pull/775

We detected that error on other models like gpt-5.4 but missed it minimax. It's an eval error not inference error. @strong-ice-84145 though ts on merging https://github.com/OpenHands/evaluation/pull/363 ? It makes rerunning eval a like this possible without running inference again