Cut to the source.The real-time AI news wire.Search tips
"exact phrase" the words together, in this order-word leave out stories with the wordtopic:robotics a topic by ID or name; quote names with spaces: topic:"open-weight models"category:"AI agents" a categorycompany:openai a company, person, product or organization (also entity:)source:forbes.com a publisher by site or nametype:analysis news, analysis, opinion, research paper and other kindsafter:2026-11-01 stories on or after that New York day; before:2026-11-10 on or before itFoundation models are general-purpose pretrained machine learning systems that serve as the underlying architecture for many downstream applications and tasks. Truncated covers their training techniques, capability improvements, hardware requirements, and regulatory developments.
AI-generated summary Researchers at the University of the Basque Country and Cardiff University tested eight AI models across 24 languages with 31,680 variations involving open-ended questions about family life, culture, and tradition. The study found that six of the eight AI models consistently selected Japan as their default answer when filling in the blanks without any provided preferences or alternatives.
AI-generated summary At the AI Engineer Code Summit, speakers discussed using Jev and other system one models for application automation and structured choices. TypeSafe AI co-founder and CEO Diogo Almeida claimed that Jev provides reliable intelligence for automating everything.
AI-generated summary BioMap announced partnerships with 11 organizations during an international summit held in Hong Kong on November 6, 2026. The initiatives encompass AI foundation-model projects spanning drug discovery, traditional Chinese medicine, agricultural breeding, and overseas market expansion.
AI-generated summary On 6 November 2026, researchers published a roadmap in Nature Machine Intelligence for building ecological foundation models trained on multiple types of ecological data at once. The roadmap outlines four areas involving data integration, ecological theory, AI architecture design, and reliability evaluation to address challenges in ecology.
AI-generated summary Google launched Gemini Decision 1, a new API backed by a multimodal decision model that scores choices in a single pass. Google says the model costs $0.10 per million input tokens with zero output token cost and handles up to 1000 questions per request.
AI-generated summary OpenAI announced on October 26th that an unreleased artificial intelligence model addressed a fundamental physics theory related to the Yang-Mills existence and mass gap problem. A committee of five external physicists evaluated the validity of the model's proof and declared themselves satisfied on November 10th.
AI-generated summary Amazon Web Services announces that DeepSeek models including DeepSeek-V4-Pro, DeepSeek-V4-Flash, and DeepSeek-V4.1-Flash are now available on Amazon SageMaker JumpStart. AWS customers can deploy these open-weight Mixture-of-Experts language models through the SageMaker JumpStart console or the SageMaker Python SDK.
AI-generated summary Anthropic is hiring researchers to teach Claude how to design physical systems and learn from industry testing. The hiring plan appears in a job posting for a Research Scientist that was live on November 11th, 2026.
AI-generated summary Déjà View is an AI model that transforms photos taken from different angles into 3D geometry and camera positions. By looping a single transformer block, the 117M-parameter model matches or outperforms models up to 10x larger across five benchmarks.
AI-generated summary Recent studies investigated whether artificial intelligence models can detect when researchers tamper with their internal neural network activity. New York University and Anthropic researcher Harvey Lederman and University of Texas linguist Kyle Mahowald tested models including Alibaba's Qwen3-235B and Meta's Llama 3.1 405B.
AI-generated summary Google is launching the Gemini Decisions API powered by the new Gemini Decision 1 model in GoogleAIStudio. The model is built for closed-set choices scoring up to 1,000 questions across text, images, audio, video, and PDFs in a single pass.
AI-generated summary Decagon introduced Alto to manage the conversational layer of its duplex voice agent by deciding when to speak and what to say. Decagon reported that Alto lifted resolution rates by more than five percentage points on live customer calls.
AI-generated summary Startup TrustedRouter introduced three encrypted combination models designed to direct workloads across multiple large language models while founder and CEO Joseph Perla stated that using a router helps maintain reliability because no single provider is completely reliable. The new offerings include Prometheus 4.0, Liberty 4.0, and Polyphemus 1.0 which are accessible through the New York Token Exchange while the company states that the models feature different routing methods, pricing, and confidential computing options.
AI-generated summary Salesforce introduced SFR-AutoR&D on November 11, 2026, stating that its agents produced SFR-Embedding v4 in about 80 hours. Salesforce reports that the research checkpoint scored 57.21 against EmbeddingGemma 2 and that production work is currently underway.
AI-generated summary Google is internally testing a new model known as Hydrogen that is designed for speed and efficiency. Staff received access to Hydrogen in the internal development platform known as Jetski, according to an internal message seen by Business Insider.
AI-generated summary @GoogleDeepMind announces AlphaGenome 2, a next-generation DNA sequence model predicting how genetic variants impact specific human cell types. @GoogleDeepMind says the model bridges single-cell biology and human genetics to help researchers uncover cellular drivers of disease.
AI-generated summary Physical AI applies learned perception, reasoning, and control to machines acting in the real world where physics and safety constraints must be managed. Multimodal systems must align signals of varying resolution and timing while separating the underlying model from the product controlling its deployment.
AI-generated summary OpenAI says its model has solved the first part of the Yang-Mills problem, which is one of the Millennium Prize Problems. David Gross and a committee appointed by OpenAI have verified and released the artificial intelligence generated proof.
AI-generated summary Anthropic unveils Claude Fable 5.2 and Claude Mythos 5.2 alongside features, speed, and price details. Meta sets India pricing for the Muse AI agent ahead of its upcoming launch.
AI-generated summary A public GitHub commit by a Google engineer's personal project mentions unannounced Gemini 4 models named Gemini 4 Hydrogen and Gemini 4 Helix. The commit maps Gemini 4 Hydrogen to the Flash tier and Gemini 4 Helix to the Flash-Lite tier.
The top AI news in your email delivered each morning. Choose categories