AI News
Lossless.shn5 Shrinks BF16 Models 1.5x to Fit a 61GB MoE on a 16GB GPU
November 14, 2026
A Hugging Face forum update reports a lossless BF16 format that’s about 1.5x smaller, and a 61 GB MoE model running on a 16 GB RTX 4080. The numbers are self-reported, but the llama.cpp comparison is unusually specific.
Read more
FlashInfer v0.7.4 Adds Rubin MoE Kernels and a Batch-Invariant Mode
November 13, 2026
FlashInfer v0.7.4 brings DeepSeek-V4 and MXFP4 MoE kernels to Rubin, plus a batch-invariant mode that returns identical bits at any batch size.
Read more
Ling-3.1-flash-1019 Reports Gains in Reasoning and Agent Benchmarks
October 27, 2026
inclusionAI’s Ling-3.1-flash-1019 reports AA-Briefcase at 1524 Elo and AutomationBench-AA at 68%, with an index score of 44. The figures are repository-reported.
Read moreAnalysis
Guides
Your Docs Chatbot Cited the Handbook and Still Got It Wrong
October 23, 2026
Your AI chatbot doesn’t memorize your files. It searches them on every question and writes from whatever it found. Here is how that works, and where it breaks.
