Alibaba has open-sourced its largest and most capable model to date, Qwen3.8-2.4T-A95B, featuring a massive 2.4 trillion parameter MoE architecture with 95 billion active parameters. This release is accompanied by the Flag OS community's achievement of Day 0 multi-chip adaptation for the model across nine different AI chip architectures, including those from Alibaba's T-Head, Nvidia, and others. The Flag OS platform enables rapid adaptation of cutting-edge models to diverse hardware, significantly reducing the time from model release to multi-chip availability and extending the utility of existing AI compute resources through optimization and quantization techniques. AI
IMPACT Accelerates deployment of cutting-edge models on diverse hardware, enhancing the value of existing AI compute infrastructure.
RANK_REASON Frontier-lab model release with system card and multi-chip adaptation details. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
