Alibaba's Qwen3.8-2.4T-A95B Drops, Autonomously Optimizes Inference Engine
BanghuaZ · x · 2026-08-12
Alibaba's Qwen team has released the weights for Qwen3.8-2.4T-A95B, a massive model with 2.4T total parameters and 95B active parameters.
Inference engine SGLang announced Day-0 support. The team demonstrated the model's strong complex engineering capabilities by having it autonomously plan and iteratively optimize SGLang's serving code. It ran unattended for 4.5 hours and achieved verified performance improvements.
More from coding & agent
- DIRECT: retrieval-based next-question suggestions for conversational agents — burkov · 2026-10-16
- Claude Code and Codex now autocomplete the user's next action — tokenbender · 2026-10-16
- Hermes Agent ships official Reverse Engineer Anything plugin with one-command setup — Teknium · 2026-10-16
- MLExpert: 35B MoE Agent Trained to Autonomously Handle Machine Learning Engineering — Shangheng Du · 2026-10-16
- Cowork-Gym: Alibaba Synthesizes 64.9K Agent Work Environments, Boosting Qwen by 8.7 Points — Anni Zou · 2026-10-16
- MMSearch-OPD: Joint RL and On-Policy Distillation for Long-Horizon Multimodal Search Agents — PekingUniversity · 2026-10-16