Alibaba launches Qwen 3.8 Max with 2.4T parameters and open weights coming next week
Latent Space · rss · 2026-08-04
Alibaba’s Qwen 3.8 Max launch centers on a new flagship model with 2.4T parameters, a 1M-token context window, and pricing of $2 input / $6 output per million tokens.
Key points:
- Alibaba says open weights for Qwen 3.8 Max and Qwen 3.8 27B are coming next week.
- The company positions it around long-horizon coding, cowork/agentic workflows, chip-design optimization, and multimodal reasoning.
- Third-party reports claim 95B active parameters, 128k output tokens, and strong benchmark results, including SWE-bench 87.3% and Terminal-Bench 2.1 at 67.4.
- Arena/Vals results put it near the top of current open models, with especially strong coding and vision performance.
The broader takeaway is that Alibaba is now pushing a very large open-weights frontier model directly into the top tier of coding and agentic workloads.
More from coding & agent
- Reverse Engineering Reveals Hidden Code in Claude Code Targeting Chinese Users — aftahi_ai · 2026-08-04
- Xberg: Local Document Reading Tool for Agents Supporting 101 Formats — Goldziher · 2026-08-04
- Hermes Agent Optimized: Boosting Small Model Efficiency via 250k Conversation Analysis — max_paperclips · 2026-08-04
- Develop with Two Prompts, Deploy with One: AI Coding Integrated with Stripe — jeff_weinstein · 2026-08-04
- Berlin AI Meetup: Framework-Free Agents and Applied NLP — hugobowne · 2026-08-04
- Qwen 3.8 Early Tests: Good at Agentic Coding, But Struggles with Long Loops — bindureddy · 2026-08-04