Qwen3-Coder-Next Technical Report

Benchmark Model Rank Results
code-generation-on-multipl-eQwen3-Coder-Next#1Accuracy: 88.23
code-generation-on-aider-polyglotQwen3-Coder-Next#2Accuracy: 66.2
code-generation-on-terminal-bench-2-0-terminus-2Qwen3-Coder-Next (Terminus2-json)#8Accuracy: 36.2
multi-task-language-understanding-on-mmlu-reduxQwen3-Coder-Next#13Accuracy: 91.18
multiple-choice-question-answering-mcqa-on-supergpqaQwen3-Coder-Next#13Accuracy: 57.45
mathematical-reasoning-on-aime24Qwen3-Coder-Next#14Acc: 89.01
code-generation-on-swe-bench-multilingualQwen3-Coder-Next (OpenHands)#27Accuracy: 64.3
mathematical-reasoning-on-aime-2025Qwen3-Coder-Next#36Accuracy: 83.07
code-generation-on-swe-bench-verifiedQwen3-Coder-Next (OpenHands)#37Accuracy: 71.3
multi-task-language-understanding-on-mmlu-proQwen3-Coder-Next#41Accuracy: 80.52
code-generation-on-livecodebench-v6Qwen3-Coder-Next#43Pass@1: 58.93