One attempted fix to the deadlock problem: @millions-crayon-32880 and @strong-ice-84145 Fix fork+threading deadlock by switching to spawn multiprocessing by neubig · Pull Request #459 · OpenHands/benchmarks

Interesting, we don't have a way to reproduce the deadlocks. but I'm testing if the happy path works with that fix here https://github.com/OpenHands/evaluation/actions/runs/22457647424

Thanks! @millions-crayon-32880 btw could you add the models you're currently working on to "expected models" so we can see how far we are from having them complete in CI on openhands-index-results?

Merging the PRs, passed the test. Also updated status in linear: ALL-3846: Support GLM-4.7 ALL-5333: Support GLM 5 ALL-5334: Support Qwen3-Coder-Next ALL-5414: Support qwen3.5-flash ALL-5335: Support claude-sonnet-4-6

Cool, OK, so those are the models we’re working on now?

@limited-student-10747 sorry for the PR review request spam. Please review them even if they are failing the precommit hooks. (I'm only fixing the metadata.json on one PR merging that, and the other PRs of that model will get it for there.) Add swe-bench results for GLM-5 Add gaia results for GLM-5 Add commit0 results for GLM-5 Add swt-bench results for GLM-5 Add gaia results for Qwen3-Coder-Next Add commit0 results for Qwen3-Coder-Next Add swt-bench results for Qwen3-Coder-Next Add swe-bench results for claude-sonnet-4-6 Add gaia results for claude-sonnet-4-6 Add swt-bench results for claude-sonnet-4-6 Add swe-bench-multimodal results for claude-sonnet-4-6 Add commit0 results for claude-sonnet-4-6 Add gaia results for glm-5

thanks, could you review my above one in return?

yes, merged!