Qwen3.8-27B

Benchmark Model Rank Results
computer-use-agents-on-osworld-verifiedQwen3.8-27B#3Success Rate: 84.3
visual-question-answering-vqa-on-realworldqaQwen3.8-27B#3Accuracy: 85.9
code-generation-on-livecodebench-v6Qwen3.8-27B#4Pass@1: 90.3
computer-use-agents-on-androidworldQwen3.8-27B#5Success Rate: 81.9
instruction-following-on-ifbenchQwen3.8-27B#6Accuracy: 79.5
code-generation-on-swe-bench-proQwen3.8-27B#10Resolved: 61.7
question-answering-on-humanity-s-last-exam-hleQwen3.8-27B#10Accuracy: 30.8
optical-character-recognition-ocr-on-omnidocbench-v1-5Qwen3.8-27B#11Accuracy: 91.1
code-generation-on-nl2repoQwen3.8-27B#15Accuracy: 42.3
multiple-choice-question-answering-mcqa-on-gpqa-diamondQwen3.8-27B#15Accuracy: 89.2
code-generation-on-terminal-bench-2-1Qwen3.8-27B (Terminus)#23Accuracy: 73.0
code-generation-on-deepsweQwen3.8-27B (Claude Code)#24Pass@1: 0.422