AI news Nov 16: Qwen4-Max and the race for AI infrastructure

Alibaba has launched Qwen4-Max with a one-million-token context window, while AMD and NVIDIA highlighted new advances in AI models, supercomputing and inference. Companies also announced fresh healthcare, robotics, security and enterprise AI initiatives.
- Alibaba launched Qwen4-Max with 2.4 trillion total parameters and a one-million-token context window.
- AMD completed its World Labs acquisition and opened Lux to Oak Ridge researchers.
- NVIDIA reported leading Vera Rubin results across training and agentic inference.
- OpenAI expanded healthcare agents, plugins and safety research.
- New announcements covered AI infrastructure, robotics, satellites and model security.
Models and inference
Alibaba released Qwen4-Max, calling it the most capable model in the Qwen family to date. It combines 2.4 trillion total parameters, native vision capabilities and a one-million-token context window for work, creative production and research tasks. Novita AI is offering the model as a serverless API with a displayed 977K-token context window and up to 393,000 output tokens.
Liquid AI released d1.2 and d1.2-ultra, adding audio input and five new API tasks. The company says d1.2-ultra ranks first among decision models on the Artificial Analysis Decision Index v2.1. Hugging Face is switching language-model routing to OpenRouter, while OpenRouter is bringing multi-provider inference, automatic failover and provider controls to the platform.
AMD and NVIDIA push AI computing forward
AMD completed its acquisition of World Labs, adding the company’s model expertise and team. Dr. Fei-Fei Li will join AMD as executive vice president and chief scientist, reporting directly to CEO Lisa Su. AMD’s Lux supercomputer opened to researchers at Oak Ridge National Laboratory for the U.S. Department of Energy’s Genesis Mission. Lux is the initiative’s first dedicated U.S. AI Factory and ranks 14th on TOP500, with 220.79 petaflops.
NVIDIA says Vera Rubin swept MLPerf Training v6.1, post-training and agentic inference, with more than 10x generational gains in agentic AI throughput per watt. Related results report 5x to 10x higher throughput per kilowatt than GB300 NVL72 in three agentic AI models and up to 2x more performance per GPU on training benchmarks.
AI infrastructure and commercial deployment
Lambda is deploying NVIDIA Photonics across a 10,000-GPU AI Factory, while a separate announcement describes 10,000+ NVIDIA Blackwell Ultra GPUs connected at its Quebec campus. CoreWeave reported multi-rack Vera Rubin MLPerf results, including a 17.65-minute DeepSeek-V3 671B run across 256 GPUs.
Dell and AMD expanded AI infrastructure from single nodes to full racks, while Dell expanded its AI Data Platform for governed enterprise data. Nokia, Nscale, Skeleton, Telia and Verda formed a Northern European AI Gigafactory consortium.
Healthcare, robotics and security
OpenAI added computer-use agents and more than 20 healthcare plugins to ChatGPT for Healthcare, alongside improved clinical search and human review controls. The company also reported healthcare results covering clinical safety, helpfulness and administrative workflows.
Pony.ai and Toyota began fully driverless bZ4X robotaxi services in Guangzhou and are working toward around 1,000 vehicles across major Chinese cities. Wiz launched an AI Security Agent to map AI application attack paths, while CrowdStrike warned about backdoored LoRA models in AI supply chains.
Other announcements
Flagship Pioneering committed £100 million to a London AI venture lab. ICEYE unveiled Gen5, a SAR satellite system with resolution as fine as 11 cm and onboard AI. ASUS plans customer pilots for its Gallop humanoid robot in the second half of 2027.
Other launches included SAP’s completed TechWolf acquisition, LanceDB’s $57.5M Series B, Apple’s US health AI study and a £1 million AI skills partnership between The King’s Trust and OpenAI.

