Hy4 preview

Benchmark Model Rank Results
agents-on-mcpatlas-public-500-tasksHy4 preview#2Accuracy: 83.7
physics-question-answering-on-critptHy4 preview#3Accuracy: 16.9
code-generation-on-nl2repoHy4 preview#4Accuracy: 58.9
agents-on-posttrainbench-v1-1Hy4 preview#5Accuracy: 35.6
code-generation-on-swe-bench-multilingualHy4 preview (swe-agent scaffold)#5Accuracy: 82.9
code-generation-on-swe-bench-proHy4 preview#5Resolved: 65.7
multiple-choice-question-answering-mcqa-on-gpqa-diamondHy4 preview#6Accuracy: 92.3
vulnerability-reproduction-on-cybergymHy4 preview#6Success Rate: 78.4
agents-on-agents-last-exam-ale-cliHy4 preview (Claude Code)#8Pass Rate: 22.8
agents-on-automationbench-publicHy4 preview#9Success Rate: 32.1
code-generation-on-terminal-bench-2-1Hy4 preview (Claude Code)#10Accuracy: 85.4
code-generation-on-programbenchHy4 preview (Claude Code)#11Almost Resolved: 17.5
code-generation-on-deepsweHy4 preview (mini-swe-agent)#15Pass@1: 0.643