Flash

TopicQwen4-Max

Flash·Models & Products·2026-11-16 13:30

Alibaba Officially Launches Qwen4-Max: 2.4T Parameters, Task Cost Down 62%

Alibaba officially launched its next-generation flagship model Qwen4-Max on November 16, disclosing its full architecture, benchmark results and API pricing: a 2.4-trillion-parameter backbone with 71 billion parameters active per token, a one-million-token context window, a 52.69 score on its in-house Work360 evaluation, and an average cost of $1.67 per task — 62% below the previous Qwen3.8-Max.

A sunlit tech office with a floating holographic interface showing an orange Q-shaped logo and streams of code
AI-generated illustration

What happened

On the morning of November 16, Alibaba's Qwen account posted “Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery!”, announcing the model's official launch; Reuters reported the same day that Alibaba had unveiled its new flagship model with strengthened safety protections. The launch comes five days after the November 11 preview debut, whose window had been scheduled to run until November 25.

According to the official blog, Qwen4-Max is positioned as a round-the-clock teammate for complex coding, professional office work and long-running agent tasks, spanning 3D modeling, game development, precision manufacturing, finance, law, healthcare and frontier scientific discovery. The model natively takes text, image and video input, with a one-million-token context window.

Alibaba said Qwen4-Max's training included security checks alongside coding tests, to reduce vulnerabilities such as code injection, memory-safety flaws and broken access controls; safeguards against misuse in cybercrime and hazardous materials were strengthened, and internal testing found no evidence of significant high-risk capabilities.

Qwen4-Max is available through the QwenCloud pay-as-you-go API, the Token Plan, Qwen Studio, QwenWork and Qoder, with interfaces compatible with OpenAI and Anthropic; the flagship's weights are not open, offered only as a hosted service, and Alibaba said smaller Qwen4 models will be open-sourced, with no date given.

Key facts

  1. Architecture: Mixture-of-experts model on the new Qwen4 architecture: a 2.4-trillion-parameter backbone, 71 billion parameters active per token, plus 328 billion N-gram embedding parameters
  2. Context: One-million-token context window, with native text, image and video input
  3. In-house benchmark: Work360 score of 52.69, up from 47.74 for Qwen3.8-Max; average cost per task of $1.67, down from $4.43, a 62% cut
  4. Token use: About 24% fewer tokens on the same work at Xhigh effort and 35% at High effort, and 48% to 49% fewer on Terminal-Bench 4.0
  5. API pricing: $1.50 per million input tokens and $4.50 per million output tokens, 25% below Qwen3.8-Max; implicit caching at $0.075 per million, down 70%
  6. Availability: QwenCloud pay-as-you-go API, Token Plan, Qwen Studio, QwenWork and Qoder; interfaces compatible with OpenAI and Anthropic
  7. Open weights: The flagship's weights are not open, offered only as a hosted service; Alibaba said smaller Qwen4 models will be open-sourced, with no date given
  8. Coding benchmarks: Bloomberg reported that Alibaba said Qwen4-Max ranks near the top of coding benchmarks, ahead of Anthropic's Claude Fable 5.1 and Moonshot's Kimi K3.1 Pro; the claim comes from the company and has not been independently verified

Why it matters

Qwen4-Max is the first Qwen flagship to launch with full architecture, benchmarks and pricing disclosed, at 62% lower cost per task than its predecessor, five days after its preview debut.
Useful Tap if this story helped you
Send to a friendMakes a share image for X, WeChat, Telegram and more

SourcesReuters (Nov 16, 2026); Dow Jones (Nov 16, 2026). This article is compiled from public information and does not constitute investment advice.

From HIMIAO Horizon · himiao.ai

Grab todayOpenAI Codex for Students: $100 in Free CreditsHIMIAO Picks →

Follow HIMIAO Horizon

No sign-up · Unsubscribe any time

Email

Notifications

A ping on your phone or computer when the brief is out. No email needed.

HIMIAO PicksWeekly freebies: one email every Friday, a 5-minute read

Comments

  1. Loading comments…
Ask the cat