Artificial intelligence news.
Straight from the original announcements.

AI news Nov 19: model tests, infrastructure and research

November 19, 2026, 21:00By the AI News Bites Editorial Team
AI news Nov 19: model tests, infrastructure and research
Image: AI News Bites

OpenAI’s GPT-6.2 Astra and Sol both completed a 3D pyramid project, but produced different strengths. Elsewhere, companies announced new AI research partnerships, infrastructure investments and tools for deploying agents in businesses.

Key points
  • GPT-6.2 Astra and Sol both built the 3D pyramid project, with different strengths.
  • Ai2 open-sourced Theorizer for scientific research.
  • MegaRobo raised nearly US$200 million for physical-world AI infrastructure.
  • Lenovo’s AI-related revenue surpassed US$10 billion.
  • Enterprise platforms added agent identities, controls and recovery tools.

Models and agent development

Kilo tested OpenAI’s GPT-6.2 Astra and GPT-6.2 Sol by asking both models to build an interactive 3D scene of workers constructing a pyramid. Both completed the project, with Astra producing the more polished visual result and Sol offering clearer operational feedback.

Google also demonstrated multi-turn agentic reinforcement learning with Gemma 4, using Tunix and a single Cloud TPU v6e-8 host to train an agent to play Plane Strike. Vercel added Liquid AI’s d1.2 and d1.2-ultra models to its AI Gateway, while Mastra launched a gateway covering more than 500 AI models.

Research and scientific AI

Ai2 launched Theorizer in Asta, an open-source AI agent that combines findings from scientific papers to propose explanations for observations in research. Researchers can examine the code and adapt its methods.

Isomorphic Labs and MSD agreed to collaborate on small-molecule drug candidates for a strategically important cancer indication. The work combines Isomorphic Labs’ AI drug design platform with MSD’s proprietary data and drug discovery expertise.

Google also detailed Streaming DiLoCo, an approach in its open-source MaxText framework for training language models across geographically distributed compute slices.

AI infrastructure and business

MegaRobo Technologies raised nearly US$200 million in a funding round involving Oman’s sovereign wealth fund, the Oman Investment Authority. The company plans to invest in core technologies, expand its AI for Science business globally and develop infrastructure connecting AI with the physical world.

Lenovo reported record second-quarter revenue of US$30.2 billion for fiscal year 2026/27, up 48% year-on-year. AI-related revenue surpassed US$10 billion, while adjusted net income rose 166% to US$1.4 billion. ASUS is ramping up NVIDIA Vera Rubin NVL72 shipments in Q4 and presenting an AI Factory reference architecture at SC26 in Chicago.

An MIT study linked US data center electricity use to distant air pollution and estimated about 650 premature deaths across the country in 2023.

Enterprise deployment and security

Microsoft Foundry now gives autonomous agents their own enterprise identities in Teams and other Microsoft 365 applications, including separate accounts, permissions, audit trails and accountable managers.

GitLab 19.6 adds agent activity controls and web search, while related updates cover scheduled engineering automations, agentic workflows in Slack and AI-assisted rotation of leaked access tokens. AWS expanded SageMaker HyperPod recovery and infrastructure checks, and Databricks said it had mitigated the Java4Shell OpenJDK vulnerability before public disclosure.

Related news

Roundups
RoundupsNovember 18, 2026, 21:00

AI news Nov 18: enterprise agents, major funding and tougher oversight

Microsoft has announced a broad set of agent-focused products and Azure services, while California is proposing stronger controls for artificial intelligence. Isomorphic…

AI news Nov 17: New frontier models and enterprise AI expansion
RoundupsNovember 17, 2026, 21:00

AI news Nov 17: New frontier models and enterprise AI expansion

OpenAI, Alibaba Cloud and Microsoft are putting new AI models in front of professional and enterprise users, with a focus on capability, cost and specialized workloads.…

AI news Nov 16: Qwen4-Max and the race for AI infrastructure
RoundupsNovember 16, 2026, 21:00

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.…