Qwen3.8 Launches With 2.4T Parameters, Tops Open-Weight Rankings
Alibaba Qwen3.8 drops with 2.4T parameters, most powerful open-weight model, second only to Anthropic Fable 5. Here is what this means for developers.
27B on a Phone: The Density Numbers Behind PrismML’s Bonsai Breakthrough
PrismML Bonsai 27B uses ternary quantization to compress a 27B model to 5.9GB, enabling phone deployment. 10x intelligence density versus full-precision. Benchmark gaps analyzed.