豆包 2.1 Pro模型更新，已接入豆包工作

TL;DR - ByteDance updated Doubao 2.1 Pro to version 0915, improving multimodal coding, long-horizon agent tasks, and tool use while reducing token consumption. The model is now broadly available through Volcano Ark’s API and integrated into Doubao Work.

• Multimodal understanding improved for video reasoning, 3D object recognition, and image-to-code workflows spanning web design, games, and 3D modeling.
• In a reported test on 1,000 historical Luanti issues, parallel sub-agents brought 83% of fixes to a merge-ready standard within nearly 36 hours.
• Long-running research and automation tasks gained stronger source tracing, multilingual retrieval, recency assessment, and cross-source data verification.
• Image and video reasoning reportedly consume over 30% fewer tokens than the previous version, with fewer reasoning rounds and tool calls.