• multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
• mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.
• quantization — Shrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.

An open-weights multimodal MoE previewing Qwen4's architecture, with firsthand notes on which Unsloth quantizations actually fit and run on a single DGX Spark.

• postQwen3.8-Flash opens up the Qwen4 architecture at 6B active parametersQwen
• articleDeepSeek V4 Pro 0813 (on OpenRouter)simonwillison.net
• articleExperiment with Qwen3.8-Flash-Next 176B Model on NVIDIA GB300 NVL72 for Agentic CodingMichelle Horton

Checking sign-in…

Loading comments…