Alibaba's Qwen3.8-Max Announced: 2.4T Parameter MoE, Open Weights Next Week
haider1 · x · 2026-08-03
Alibaba's Qwen team has announced that it will release the open weights for its most capable model to date, Qwen3.8-Max, alongside Qwen3.8-27B next week.
Key highlights of Qwen3.8-Max:
- Architecture & Scale: A 2.4T-parameter Mixture-of-Experts (MoE) model.
- Autonomous Coding: Capable of 10+ days of self-evolving development, achieving end-to-end autonomous closure from an empty folder to a production-ready project.
- Long-Horizon Planning: Features system-level autonomous planning and closed-loop adaptive learning, enabling tasks like 500+ turns of chip design optimization and 365 days of e-commerce strategy.
- Pricing: Highly competitive API pricing at $2/M input tokens and $6/M output tokens.
According to the reposter, the model matches or surpasses closed-source competitors like GPT-5.6 Sol and Opus 4.8 across multiple benchmarks.
Related event: Alibaba releases Qwen3.8-Max, open-sources weights next week(44 posts)→
More from Models
- Users report Gemini's memories can't actually be deleted even after wiping chat history — ThrowRA_Purpose2025 · 2026-10-16
- Unverified: Gemini 4 Argon reportedly scores 97.5 on ARC-AGI 3 as benchmark saturates — Different_Ad1136 · 2026-10-16
- Gemini users report usage limits suddenly and drastically reduced — SemFi · 2026-10-16
- OpenAI stealth-testing mysterious internal model 'Aeon' in Codex desktop app and Dots — cedric_chee · 2026-10-16
- Argon edges out astra as daily driver, but Claude still wins heavy coding work — iruletheworldmo · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16