OpenAI's internal Astra model has generated solutions to ten long-standing open problems across mathematics and theoretical computer science, with the arguments formalized in Lean. These results span high-dimensional geometry, coding theory, group theory, and quantum complexity, addressing questions that had seen no progress for at least a decade.
Google has introduced Gemini Robotics 2, an intelligence layer designed to provide adaptable robots with intelligent whole-body control, advanced dexterity, and the ability to collaborate in teams. This update expands physical AI capabilities beyond narrow, repetitive tasks by enabling robots to reason through movements and adapt to new bodies quickly.
GPT-5.6 enhances artificial intelligence efficiency across multiple dimensions, including model architecture, inference processes, and agentic workflows.
Google has introduced three new models in its Flash series: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber, designed to improve token efficiency, latency, and reliability for production AI agents.
Anthropic has released Claude Opus 5, a new model designed to power long-running agents while delivering significant improvements in coding and professional work. It is positioned as a step-change improvement for the Opus tier, offering performance close to Claude Fable 5 at half the cost.
Anthropic has paused internal activities involving its upcoming model, Astra, because recent evaluations indicate it may possess critical cybersecurity capabilities under the company's Preparedness Framework. The firm cannot rule out that the model can identify and develop functional zero-day exploits in hardened real-world systems without human intervention.
OpenAI has released GPT-Live, a third-generation voice system that eliminates the traditional turn detector by using a full-duplex model capable of listening and speaking simultaneously. This architecture allows for more immediate and natural conversations while maintaining low latency through a new system design optimized for real-time performance.
Microsoft Research has released Orchard, an open-source framework designed to enable scalable and cost-effective research into autonomous agents. The system centers on Orchard Env, a Kubernetes-based environment service that allows researchers to train and evaluate agents across diverse domains like software engineering, web navigation, and personal assistance without rebuilding infrastructure.
OpenAI has announced significant price reductions for its GPT-5.6 model family, cutting the cost of GPT-5.6 Luna by 80 percent and GPT-5.6 Terra by 20 percent. The company frames these changes as part of a broader strategy to make advanced AI more affordable and widely useful through improved infrastructure efficiency.
DeepSeek has launched the official public beta of the DeepSeek-V4-Flash API, featuring significantly enhanced agent capabilities compared to the V4-Pro-Preview. The update involves re-post-training the model while maintaining its original architecture and size.
Microsoft Research has released Echoverse, a collection of twelve deep synthetic environments designed to train computer-use agents. The initiative includes ten domain-specific worlds and two capability-focused worlds, all built to maintain coherent state and behavioral fidelity.
Google has launched Gemini Robotics ER 2, a new embodied reasoning model designed to help robots think fast and plan multi-step tasks in real-time. The update introduces significant upgrades over the previous ER 1.6 version, including continuous video-based progress tracking, precision moment-finding, and native multi-robot collaboration capabilities.
xAI has introduced Build Mode, a new feature available in Early Beta for SuperGrok Heavy subscribers that allows users to create websites, apps, games, and interactive dashboards directly within the Grok chat interface.
Grok Build now supports workflows, allowing users to describe complex tasks in plain language that are automatically planned and executed across hundreds of parallel agents. These orchestration scripts fan out work, verify results, and report back in a single background run while keeping the user's session free.
OpenAI has introduced "Presence," a new product designed to help enterprises deploy reliable AI agents capable of handling high-value production workflows such as customer support and internal IT requests. The system combines model reasoning with specific policies, guardrails, and escalation rules to ensure accuracy and control.
Google has introduced Gemini 3.5 Flash Cyber, a lightweight cybersecurity model built on top of the 3.5 Flash architecture and fine-tuned to find, validate, and patch software vulnerabilities more efficiently than mainline Flash models.
Google DeepMind and Isomorphic Labs have outlined their joint strategy to enhance global biosecurity by leveraging AI to prevent misuse, detect outbreaks, and respond to biological threats. The companies emphasize the need for frontier AI models to help society build resilience against future pandemics and safety risks.
SpaceXAI has released the source code for Grok Build, its coding agent and terminal user interface (TUI), on GitHub. This move aims to provide a robust and reliable harness by allowing users to inspect exactly how context assembly and tool-call dispatch work.
OpenAI provides guidance on evaluating AI usage and controlling spend as teams transition from simple chat to longer-running, multi-step workflows. The article emphasizes that leaders must look beyond token prices to measure useful work per dollar, such as tasks completed or time saved.
Anthropic has introduced Claude for Teachers, a free tool providing verified K-12 educators in the US with premium Claude capabilities and access to evidence-based curricula. The platform connects to Learning Commons to align lesson planning with academic standards across all 50 states and integrates with various educational tools like ASSISTments, Brisk Teaching, and Canva Education.