Qwen3.8-Max Announced with 2.4T Parameters, Open-Sourcing Next Week
xiaohu · x · 2026-08-03
Qwen3.8-Max has been announced and is slated for open-source release next week, marking the first open-source model in the Qwen Max series. It features a massive 2.4 trillion total parameters and 95B activated parameters.
The model focuses on long-cycle autonomous work and self-evolution, showcasing impressive results in extreme tests:
- Autonomous Coding: Ran completely autonomously for 16 days in the oh-my-cli project.
- Beating Humans: Autonomously built and tuned ensemble models for an e-commerce multimodal intent recognition competition, beating 87% of human teams.
- Simulated E-commerce: Navigated real-world market fluctuations and fraud traps to achieve a 4.16x return on investment.
- Research Reproduction: Spent 5 days writing 7,600 lines of code to reproduce a recent paper, designing 18 improvements to beat the original benchmark.
Additionally, it supports processing 100+ hour videos and features a visual feedback loop, allowing it to review its own generated UI layouts and autonomously correct errors.
Related event: Alibaba releases Qwen3.8-Max, open-sources weights next week(44 posts)→
More from coding & agent
- DeepMind turns coding agent Antigravity into a science tool, surfacing novel antimicrobial peptide in one session — pushmeet · 2026-10-16
- Cross-review coding workflow: Claude writes the plan, Codex audits it — theteknosaur · 2026-10-16
- Simon Willison: code mode is the future of tool calling, if it can be sandboxed — springrod · 2026-10-16
- Normalizing Claude's schema-free MCP output before it hits downstream n8n nodes — snowingbol · 2026-10-16
- Agent named 'Ilya' reportedly scores 1.1pt higher on benchmarks — daniel_mac8 · 2026-10-16
- A plain notes.md fixed my agent's cross-session amnesia better than chat history — rahul_watertech · 2026-10-16