LongCat-Flash Technical Report

Benchmark Model Rank Results
instruction-following-on-arena-hard-auto-v2LongCat-Flash#1Accuracy: 86.50
logic-puzzle-solving-on-zebralogicLongCat-Flash#1Accuracy: 89.30
agents-on-bench-telecomLongCat-Flash#6Accuracy: 73.68
multi-task-language-understanding-on-c-evalLongCat-Flash#8Accuracy: 90.44
mathematical-reasoning-on-math-500LongCat-Flash#18Accuracy: 96.40
multi-task-language-understanding-on-mmlu-proLongCat-Flash#30Accuracy: 82.68
mathematical-reasoning-on-aime24LongCat-Flash#47Acc: 70.42
code-generation-on-livecodebench-v6LongCat-Flash#53Pass@1: 48.02
code-generation-on-swe-bench-verifiedLongCat-Flash (OpenHands)#64Accuracy: 60.40
mathematical-reasoning-on-aime-2025LongCat-Flash#67Accuracy: 61.25
multiple-choice-question-answering-mcqa-on-gpqa-diamondLongCat-Flash#75Accuracy: 73.23