Qwen/Qwen3.8-2.4T-A95B

Timeline

1. Oct 17, 2026 · Tutorial / explainer · 14 sources
   Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio
   In this post, we show how to run the latest frontier-class model — Muse Spark Open, with nearly a trillion parameters — right from your desk, using the latest version of MLX, released today.
2. Oct 12, 2026 · Research paper · 2 sources
   Alibaba Qwen Introduces QwenGyre: Elastic Reinforcement Learning RL for XLong-Horizon Agents
   Alibaba’s Qwen team has introduced QwenGyre, an elastic reinforcement learning RL framework for extreme-long-horizon agent training.

Often appears with