Inspur Open-Sources Trillion-Parameter Multimodal Large Model, Significantly Improving Computing Efficiency
On March 9, 2026, Inspur Information released and open-sourced a trillion-parameter multimodal large model adopting the Mixture of Experts (MoE) architecture. The model boosts pre-training computing efficiency by 49%, inference speed by 35%, and reduces costs by 40%. It excels in enterprise scenarios like document understanding, table analysis, and multimodal interaction, driving domestic large models from labs to large-scale industrial deployment.
Keywords: Inspur Information, Trillion Parameters, Multimodal Large Model, MoE Architecture, Computing Efficiency
new
ChatGPT, other AI chatbots approved for official use in US Senate, NYT reports
AI startup Thinking Machines clinches capital and a major chip supply deal from Nvidia
ChatGPT, other AI chatbots approved for official use in US Senate, NYT reports
World's largest e-commerce platform: Amazon ventures into AI healthcare.
Meta Acquires Moltbook, the Social Network Just for A.I. Bots
Inspur Open-Sources Trillion-Parameter Multimodal Large Model, Significantly Improving Computing Efficiency
OpenClaw 3.8 Released Rapidly, Global Developers Launch "Lobster Raising" All-Night Craze
Tencent Quietly Beta Tests "QClaw": A One-Click Deployment Tool Based on OpenClaw Framework, Lowering Agent Development Barriers
Anthropic’s Claude found 22 vulnerabilities in Firefox over two weeks
OpenAI delays ChatGPT’s ‘adult mode’ again