DiffusionGemma Technical Report

Benchmark Model Rank Results
long-context-understanding-on-mrcr-v2-128kDiffusionGemma 26B A4B#5Accuracy: 32.0
multi-task-language-understanding-on-mmmluDiffusionGemma 26B A4B#10Accuracy: 81.5
multimodal-reasoning-on-mathvisionDiffusionGemma 26B A4B#17Accuracy: 70.5
agents-on-benchDiffusionGemma 26B A4B#25Accuracy: 56.2
code-generation-on-livecodebench-v6DiffusionGemma 26B A4B#34Pass@1: 69.1
multimodal-reasoning-on-mmmu-proDiffusionGemma 26B A4B#38Accuracy: 54.3
question-answering-on-humanity-s-last-exam-hleDiffusionGemma 26B A4B#39Accuracy: 11.0
mathematical-reasoning-on-aime-2026DiffusionGemma 26B A4B#42Accuracy: 69.1
multi-task-language-understanding-on-mmlu-proDiffusionGemma 26B A4B#48Accuracy: 77.6
multiple-choice-question-answering-mcqa-on-gpqa-diamondDiffusionGemma 26B A4B#76Accuracy: 73.2