Google DeepMind releases AlphaGenome 2 model
Google DeepMind has unveiled the AlphaGenome 2 model, designed to predict how genetic variants impact gene expression within specific human cell types and states. The new model builds on data from more than 100 million single-cell profiles across 68 human tissues to support rare disease research and complex genetic traits.
Google DeepMind formally verifies Signal’s encryption mathematically
Google DeepMind has announced AlphaProof Code, a system designed to transform software security specifications and implementation details into machine-checkable mathematical proof obligations. The team applied the new system to the cryptographic core of the Signal messaging protocol, generating and checking millions of lines of Lean language proofs over the course of several weeks.
Google launches Gemini Omni model for video generation
Google rolled out the Gemini Omni 1.2 Flash video generation model for general availability on Thursday, November 5, 2026, extending generated clip lengths to 15 seconds instead of 10 seconds at a 30% lower cost than the previous version. The new model supports multi-turn conversational editing with high character and color consistency and precise lip-syncing.
Google DeepMind unveils Genie 4 audio world model
Google DeepMind announced on Tuesday, October 27, the launch of Genie 4, the new generation of its AI world model that generates interactive real-time virtual environments with unprecedented realism, featuring native capabilities to produce music, ambient sounds, and dialogue.
As language models slow down, AI giants bet on “world models” to understand the physical reality
In a major strategic shift, companies like Google, Meta, and NVIDIA are stepping up their investments in “world models,” a technology aimed at understanding and interacting with the physical world through video and robotics data, amid signs that large language models have hit a development ceiling.
Boston Dynamics aims to build Atlas in one hour
Boston Dynamics has published a short film from its headquarters in Massachusetts revealing its plan to cut the assembly and build time for the humanoid robot “Atlas” from a full day currently to just one hour next year, and eventually to a few minutes. This manufacturing acceleration comes after its entire 2026 production run was booked by Hyundai Motor Group and Google DeepMind.
DeepMind solves nine mathematical problems
Google DeepMind has published a peer-reviewed study in the journal Science demonstrating the success of its AlphaProof Nexus system in solving nine open problems posed by mathematician Paul Erdős. The system combines Gemini models with the Lean theorem assistant for automated proof verification.
Google DeepMind revolutionizes hurricane forecasting
Google DeepMind’s new AI hurricane model outperforms traditional models and even human forecasters, delivering faster and more accurate predictions that could save lives and property.
Google launches Gemini 3: A leap toward artificial superintelligence
Google has announced the launch of its newest and smartest model, Gemini 3, which ranks as the world’s best in multimodal understanding and coding. Discover the capabilities of Gemini 3 Pro and the new Deep Think mode, which pushes the boundaries of artificial intelligence to new heights with an unprecedented visual and interactive experience.
Google DeepMind bolsters its ambitions in the world of robotics
In a move highlighting Google’s seriousness about developing advanced robotics, Google DeepMind has appointed Aaron Saunders, former chief technology officer of the renowned Boston Dynamics, to lead hardware engineering with the goal of making the Gemini model the primary operating system for future robots.