dots3-note Preview: A Small but Mighty Step Toward Long-Horizon Agency in Real Life

Benchmark Model Rank Results
audio-understanding-reasoning-on-mmau-prodots3-note Preview#1MMAU-Pro Overall Weighted Performance: 69.79
deep-research-agents-on-browsecomp-zhdots3-note Preview#1Accuracy: 75.78
chart-question-answering-on-charxiv-rqdots3-note Preview#2Accuracy: 83.1
multimodal-reasoning-on-mathvisiondots3-note Preview#2Accuracy: 87.66
video-question-answering-on-video-mmmudots3-note Preview#2Accuracy: 86.78
multimodal-reasoning-on-babyvisiondots3-note Preview#3Accuracy: 50
multimodal-reasoning-on-mmmu-prodots3-note Preview#3Accuracy: 79.08
video-question-answering-on-mmvudots3-note Preview#5Accuracy: 79.9
code-generation-on-nl2repodots3-note Preview#9Accuracy: 49.82
deep-research-agents-on-browsecompdots3-note Preview (w/ context management)#10Accuracy: 83.3
code-generation-on-swe-bench-verifieddots3-note Preview (live-swe-agent)#13Accuracy: 78.4
code-generation-on-swe-bench-multilingualdots3-note Preview (live-swe-agent)#15Accuracy: 75.67