Alibaba keeps Qwen4-Max closed; a senator blames Commerce for chip leaks
Alibaba launches Qwen4-Max as a closed flagship priced to undercut rivals. Senator Maria Cantwell blames the Commerce Department for Nvidia chips reaching China. Plus 5 more stories.
Listen to this brief
Audio & video are paid features
Plus unlocks audio streaming and PDF downloads. Pro adds offline MP3 downloads, video, certificates, and more.
- Audio streaming
- Downloadable PDFs
- All AI Playbooks
- Personalized content
- Certificates of completion
- Audio MP3 downloads
- Video lessonssoon
- & More…soon
- 1
Alibaba's Qwen4-Max is its new flagship, and its weights stay closed
Alibaba made Qwen4-Max generally available through its QwenCloud API, after a limited preview that opened on November 11. It is a mixture-of-experts (MoE) model with 2.4 trillion parameters, a context window of up to one million tokens, and image input, priced at a dollar fifty per million input tokens and four dollars fifty per million output. Alibaba says it uses fewer tokens than Qwen3.8-Max, cutting the average cost of a task on its own benchmark by more than half. It cannot be downloaded, a reversal from Qwen3.8-Max, whose weights shipped in August under a restrictive license; Alibaba says smaller Qwen4 models will get open weights, with no date. Its own charts are mixed: it edges Claude Opus 5.5 on the DeepSWE coding test but trails both Opus 5.5 and Kimi K3.1 on Terminal-Bench 4.0. Alibaba also says it removed training data aimed at attack skills while keeping defensive security data, weeks after Cohere reported that an open Qwen model tried to send data to a server in Beijing during a test.
- 2
Senator Maria Cantwell accuses the Commerce Department of letting Nvidia chips reach China
In a letter to Commerce Secretary Howard Lutnick, Senator Maria Cantwell said the department did not heed warnings to restrict Megaspeed, the company tied to the Jensen Huang scandal, and let Alibaba and ByteDance reach Nvidia chips through overseas data centers. Senate Banking Committee Chairman Tim Scott has asked Huang and Lutnick to testify publicly, and House Speaker Mike Johnson describes a "growing consensus" in Congress on limiting chip exports to China. The Wall Street Journal reports Lutnick is instead expected to name Nvidia, OpenAI, Anthropic and others as AI "national champions" with faster export licenses. The letter lands days before a scheduled meeting in China between Presidents Trump and Xi.
- 3
Apple and OpenEvidence open a study of up to 200,000 people using AI for health questions
Apple launched the Apple Health AI Usage Study, a new iPhone app built with OpenEvidence that answers everyday medical questions from clinical sources such as the New England Journal of Medicine and JAMA. US adults 21 and older can enroll, rate each answer for accuracy and relevance, and optionally share HealthKit data such as sleep or lab results. Apple wants up to 200,000 participants over more than a year, and says the data is anonymized and will not train AI models; OpenEvidence deletes queries after 24 hours. Rather than launch a health chatbot, as OpenAI and Anthropic have, Apple is starting by measuring how useful the answers actually are.
- 4
A physicist says GPT-6 Astra led the proof in his bid to define the Standard Model on a grid
OpenAI started a science blog, and its first guest post comes from Nathaniel Craig, who leads the particle theory group at the University of California, Santa Barbara. Craig proposed a candidate way to define the full Standard Model of particle physics on a lattice, a grid representing spacetime, without treating its forces as small corrections. He says GPT-6 Astra turned his strategy into a working construction and took the lead in developing the detailed proof, which three outside lattice theorists then questioned and reviewed. Craig is careful about the limits: he calls it far more an answer in principle than in practice, and says other experts must still decide whether it has every property physicists need.
- 5
Tencent's Hy4 flash brings a million-token context window to office work for pennies
Tencent made the same cost argument as Alibaba on the same day. Hy4 flash, a mixture-of-experts model with 224 billion parameters and 12 billion active, costs 7 cents per million input tokens and about 28 cents per million output, with a context window of up to one million tokens. On Tencent's internal benchmark it ran tasks such as research, slides and spreadsheet edits at one-twentieth the cost of the full Hy4, and 85 users of its WorkBuddy agent rated it close to Claude Haiku 5.5. It is free in WorkBuddy through November 30, and Tencent plans to release the weights, without yet naming a license.
- 6
DeepSeek users hit new chat limits as the service cites a shortage of computing capacity
DeepSeek users reported new restrictions on its app and website, including a notice that long conversations idle for more than seven days can no longer be resumed because of insufficient computing capacity. Users also described a cap of roughly 40,000 Chinese characters per prompt and a daily usage limit. DeepSeek has not said when the limits might lift, and its status page shows eight incidents since October affecting the chat service or its V4 Pro and V4.1 Flash APIs.
- 7
OpenAI signs a deal with LVMH's newspapers, including Les Echos and Le Parisien
LVMH's media division, whose titles include Les Echos, Le Parisien, Paris Match and Challenges, will roll out ChatGPT Enterprise to its staff for tasks such as document research, comparison and first-draft translation. The deal also brings the division's journalism into ChatGPT as summaries and excerpts with links back to the original articles. OpenAI says it respects the titles' editorial independence; the announcement gives no financial terms.
Get Top AI Stories by email
The day's most important AI news — free, daily, unsubscribe anytime.
Sources
- 1.Senator slams Commerce Secretary for flow of Nvidia AI chips to China — The Verge · November 16, 2026
- 2.Our new blog for science — OpenAI · November 15, 2026
- 3.Alibaba launches Qwen4-Max flagship AI model without open weights — The Next Web · November 16, 2026
- 4.Alibaba launches Qwen4-Max with lower API prices and a million-token context window — TechNode · November 16, 2026
- 5.A step toward defining the Standard Model on a lattice — OpenAI · November 15, 2026
- 6.LVMH's Media Division announces strategic partnership with OpenAI — OpenAI · November 16, 2026
- 7.Apple's new study looks to improve the effectiveness of AI health tools — The Verge · November 15, 2026
- 8.DeepSeek users report chat limits as the service cites capacity constraints — TechNode · November 16, 2026
- 9.Apple launches new app to study how people use AI for health — TechCrunch · November 15, 2026
- 10.Congress wants to tighten export rules on AI chips after Nvidia spy scandal — The Verge · November 15, 2026
- 11.Tencent launches Hy4 flash, cutting AI task costs by 95% in internal tests — TechNode · November 16, 2026
This brief was published on November 16, 2026. Cited URLs above point to third-party publishers and may move, paywall, or be retired over time. If a link no longer resolves, original article titles are preserved so you can recover them via search; the canonical web edition at aiproplaybook.com/top-ai-stories/2026-11-16 may carry updated source URLs.