Qwen3.8-Flash-Next: Inside Alibaba's Preview of the Qwen4 Architecture
Alibaba open-weighted a working preview of Qwen4's architecture: 125B parameters, only 6B active per token. It beats prior open Qwen models ...
Aug 27, 2026
6 min
Qwen3.8-Max: Alibaba Just Open-Sourced a 2.4 Trillion Parameter Monster — And It Changes Everything
Alibaba just dropped the weights for Qwen3.8-Max — a 2.4 trillion parameter Mixture-of-Experts model that autonomously coded for 16 days str...
How Chinese Models Actually Took Over OpenRouter, Month by Month
Two numbers—1.2% and 46%—capture the same 18-month rise of Chinese open-weight models on OpenRouter, but each reflects different evidence. T...
Chinese Open-Weight Models Just Won OpenRouter's Token War. They Still Haven't Won the Dollar War.
Chinese open-weight models like DeepSeek, GLM, Kimi, and Qwen now account for 30-46% of US company traffic on OpenRouter, and up to 61% in n...
Your AI, Your Rules: How to Run Powerful AI Models Locally on Your Computer in 2026
Every prompt you've ever sent to ChatGPT, Claude, or Gemini went to someone else's server. In 2026, you don't have to accept that trade-off ...
