Qwen (Alibaba) · 2026-05-20 · major
Qwen3.7-Max — Alibaba's New Agentic Flagship Runs 35-Hour Autonomous Tasks With 1,000+ Tool Calls
Alibaba unveiled Qwen3.7-Max at the 2026 Cloud Summit in Hangzhou. Marketed as a base for AI agents, the closed-weights model sits at #13 on Arena Text and #6 worldwide for Alibaba lab's text research.
Alibaba's new flagship Qwen3.7-Max is built to run agents for tens of hours and thousands of tool calls without stalling.
Key specs
| Arena text rank | #13 globally |
|---|---|
| Arena vision rank | #16 globally (Plus-Preview) |
| Arena math rank | #7 globally |
| Arena code rank | #10 globally |
| Autonomous run length | 35 hours |
| Tool calls per task | 1,000+ |
What is it?
Qwen3.7-Max is the next-generation flagship in the Qwen3.7 series, formally launched by Alibaba's Qwen team on May 20, 2026 at the Alibaba Cloud Summit in Hangzhou. Closed-weights, API-only. Preview versions of Max and Plus first appeared on chat.qwen.ai earlier in the week with deep-thinking mode enabled and web-search and code-interpreter temporarily disabled. Alibaba pitches it as a foundation specifically tuned for the 'agentic era' — long-horizon coding, office workflow automation, and multi-step reasoning.
How does it work?
Alibaba says Qwen3.7-Max can autonomously execute long-horizon agentic tasks 'sustaining continuous operation for up to 35 hours and managing over 1,000 tool calls without performance degradation', with reasoning speed roughly 10x its predecessor. Independent Arena scores currently place Max-Preview #13 globally on the text leaderboard (#1 among Chinese labs), and the Plus-Preview vision variant #16 globally. The model ships alongside an updated Alibaba Cloud stack — the Panjiu AL128 supernode (128 accelerators per rack, petabyte-per-second bandwidth) and the new T-Head Zhenwu M890 AI chip with 144 GB GPU memory and FP32-to-FP4 precision support.
Why does it matter?
Qwen3.7-Max is positioned as the agentic counterpart to GPT-5.5, Claude Opus 4.7, and Gemini 3.5 — but it lives behind the Alibaba Cloud Bailian API and was tuned for cross-day agent loops rather than chat. Combined with the Zhenwu M890 launch, it is Alibaba's clearest pitch that China can run the full 'AI factory' stack (chip, infra, model, agent platform) without leaning on US silicon. Expect it to set the bar for Chinese frontier closed models for the rest of Q2.
Who is it for?
Teams building long-running coding or workflow agents on Alibaba Cloud, and anyone benchmarking Chinese frontier models against OpenAI/Anthropic/Google.
Try it
Open chat.qwen.ai and pick 'Qwen3.7-Max-Preview' from the model menu (deep-thinking mode on by default).