| Edition | Wave | Rank | Level | Change | Title | Sources | Links |
|---|---|---|---|---|---|---|---|
| 2026-11-16 | 156 | #10 | 🔥Must-read | First seen | Qwen4-Max formally launches: 2.4T backbone, 71B activated per token, cost per task down 62% | Qwen 官方博客, Hacker News, OpenRouter, AIsa 摘要 | qwen.ai · openrouter.ai |
After the November 11 preview, Qwen published its formal blog post today: a new Qwen4 architecture with a 2.4-trillion-parameter backbone, 71B activated per token, native vision, and a 1M-token context window; it self-reports 52.69 on Work360 (vs 47.74 for Qwen3.8-Max), cost per task down from $4.43 to $1.67 (−62%), and about 48%–49% less token usage on Terminal-Bench 4.0; it reached the HN front page with 457 points / 178 comments; OpenRouter listed qwen/qwen4-max today; per the AIsa digest, Qwen also teased smaller open Qwen4 models coming soon, with Qwen4-27B able to run on 16GB of memory.
Suggested action:Write a tweet / dig deeper — a new flagship plus an open-source path is worth a full piece.