StudioinclusionAI releases Ling-3.1-flash-1019: 560B MoE with 1M context and improved reasoning and agentic performance
Original title: inclusionAI/Ling-3.1-flash-1019 · Hugging Face
inclusionAI released Ling-3.1-flash-1019, an upgrade of Ling-3.1-flash with 560B total parameters (25B activated), a hybrid-linear MoE architecture, and 1M context. Post-training improvements raise AA-Briefcase from about 1400 to 1524 Elo, AutomationBench-AA from 62% to 68%, CritPt from 18% to 26%, SciCode from 54% to 57%, and the Artificial Analysis Intelligence Index v4.3.2 score from 41 to 44. SGLang offers day-0 support with MTP (NEXTN) speculative decoding and Hopper decode optimizations, vLLM deployment instructions are provided, and a training-content summary PDF is published.
The model card gives direct benchmark comparisons with the previous version plus ready-to-run SGLang and vLLM commands, letting readers gauge the upgrade and get started quickly.
The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.