LIVE
Pedestal AI Secures Seed Funding to Automate Logistics With Intelligent Agents United Launch Alliance Investigates Vulcan Rocket Failure US Southwest Hydropower Output Hits Record Lows Amid Persistent Drought YouTube Updates View Metrics to Include TV Co-Viewing Estimates OpenAI Partners with Chilean Ministry to Advance Public Sector AI Ambience Healthcare Expands Clinical AI Suite for Hospital Operations AI Models Vulnerable to Manipulation in Simulated Warfare Scenarios, Research Shows AWS Highlights Enterprise Governance and AI Workspace Quick Ahead of re:Invent 2026 LoanPro Integrates Amazon Bedrock to Automate Loan Servicing Workflows Google Brings Wallet App to iOS Users Across the United States Pedestal AI Secures Seed Funding to Automate Logistics With Intelligent Agents United Launch Alliance Investigates Vulcan Rocket Failure US Southwest Hydropower Output Hits Record Lows Amid Persistent Drought YouTube Updates View Metrics to Include TV Co-Viewing Estimates OpenAI Partners with Chilean Ministry to Advance Public Sector AI Ambience Healthcare Expands Clinical AI Suite for Hospital Operations AI Models Vulnerable to Manipulation in Simulated Warfare Scenarios, Research Shows AWS Highlights Enterprise Governance and AI Workspace Quick Ahead of re:Invent 2026 LoanPro Integrates Amazon Bedrock to Automate Loan Servicing Workflows Google Brings Wallet App to iOS Users Across the United States

Alibaba Introduces Qwen4-Max with 2.4 Trillion Parameters and 1M-Token Context

🤖 AI Models 2 min read 👁 6

Alibaba has launched Qwen4-Max, its first generation-four multimodal model featuring a 2.4-trillion-parameter sparse Mixture of Experts architecture and a 1-million-token context window.

🤖AI Models

Key points

  • Qwen4-Max features a 2.4T-parameter backbone with 71B active parameters per token.
  • The model supports a 1-million-token context window and handles text, images, and video.
  • It is hosted exclusively through Alibaba's cloud platforms with no open-weight release.

Alibaba's Qwen team has unveiled Qwen4-Max, marking the debut of the company's Qwen4 model generation. Built as a proprietary sparse Mixture of Experts system, the model features a 2.4-trillion-parameter backbone alongside 71 billion active parameters per token, supplemented by 328 billion N-gram Embedding parameters. It handles text, images, and video while supporting an expansive 1-million-token context window and a maximum output of 393,000 tokens.

Unlike open-weight predecessors, Qwen4-Max is available exclusively via hosted enterprise channels including the QwenCloud API, Token Plan, Qwen Studio, QwenWork, and Qoder. Alongside the launch, Alibaba reduced its API pricing, claiming a 62% reduction in average task costs. Architectural enhancements include Gated DeltaNet combined with Qwen Sparse Attention, Gated Residual branches, and specialized N-gram embeddings designed to improve stability and efficiency during long-horizon professional workflows.

In benchmark evaluations provided by Alibaba, Qwen4-Max secured top positions in several specialized categories, such as legal agent work on HLAB-Vals-1.1 and the Agents' Last Exam v1.1. However, the model trails competitors like Claude Opus 5.5 and Kimi K3.1 in traditional coding benchmarks and vision tasks.

What to watch next: Enterprise adoption rates of Alibaba's new hosted platform and independent validation of its benchmark claims.

Source MarkTechPost — marktechpost.com/2026/11/16/alibaba-qwen4-max-release/ ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.