Google DeepMind

A startup founded by a former Google DeepMind researcher has secured a record-breaking $1.1 billion in seed funding. The capital is intended to support the company's mission of developing artificial superintelligence.

David Silver, the creator of AlphaGo, has founded Ineffable Intelligence to develop superintelligent AI through reinforcement learning. He argues that current LLM-based approaches are limited by human data, advocating instead for systems that can learn autonomously through trial and error.

Google DeepMind has announced a partnership with South Korea's Ministry of Science and ICT to support the nation's AI strategy and scientific research. The collaboration aims to leverage frontier AI models to accelerate breakthroughs in fields such as life sciences, weather, and climate.

Startup 10x Science has announced a $4.8 million seed round to address the bottleneck in characterizing AI-generated drug candidates. The company focuses on bridging the gap between AI-driven protein predictions and practical biological testing.

Google DeepMind has introduced Decoupled DiLoCo, a new distributed architecture designed to train large language models across geographically distant data centers. This method uses asynchronous data flow to overcome the synchronization and bandwidth challenges typically associated with large-scale, tightly coupled training systems.

Google DeepMind is partnering with major consulting firms like Accenture, BCG, and McKinsey to accelerate the adoption of frontier AI in the enterprise sector. The initiative aims to bridge the gap between AI research and large-scale industrial production through agentic transformation.

Google DeepMind has announced the Gemini Robotics-ER 1.6 model, which enhances the ability of robots like Boston Dynamics' Spot to perform visual inspections. The model enables robots to accurately read analog gauges, thermometers, and other complex instruments in industrial environments.

Google has introduced Gemini 3.1 Flash TTS, a new text-to-speech model designed for high expressivity and controllability. The model features native multi-speaker dialogue, support for over 70 languages, and a new audio tag system for granular control over vocal style.

Google DeepMind has introduced Gemini Robotics-ER 1.6, a model designed to enhance embodied reasoning in robots. The upgrade focuses on improving spatial understanding, task planning, and the ability to interpret physical objects like pressure gauges.

Google DeepMind has introduced Gemma 4, a new family of open-weights models designed for advanced reasoning and agentic workflows. The release includes four model sizes ranging from 2B to 31B parameters, optimized for high intelligence-per-parameter and mobile-first AI applications.

Google DeepMind has introduced Gemini 3.1 Flash Live, a new audio-centric model designed for high-speed, natural real-time dialogue. The model features improved reasoning, better instruction following, and enhanced tonal understanding for developers and enterprise users.

Google DeepMind has released a new toolkit and research findings designed to measure how AI models can be used for harmful manipulation. The study focuses on the ability of AI to deceptively alter human thought and behavior in controlled settings.

Google has introduced Lyria 3 Pro, an advanced music generation model capable of creating tracks up to three minutes long with better structural control. The model is being integrated into Vertex AI and Google AI Studio to support developers and businesses in scaling high-fidelity audio production.

Google DeepMind has released a new paper proposing a cognitive taxonomy to measure progress toward Artificial General Intelligence (AGI). The framework uses insights from cognitive science to identify ten key abilities, such as perception and memory, and includes a Kaggle hackathon to develop practical evaluations.

Google has introduced Gemini 3.1 Flash-Lite, a high-speed and cost-efficient model designed for high-volume developer workloads. The model offers significantly faster response times and improved performance in reasoning and multimodal benchmarks compared to its predecessor.

Google has introduced Nano Banana 2, a new state-of-the-art image model that combines advanced reasoning and world knowledge with high-speed performance. The model aims to make professional-grade features like precision text rendering and complex data visualization accessible to a broader audience.

Google has announced the release of Gemini 3.1 Pro, an upgraded model designed for advanced reasoning and complex problem-solving. The model is being rolled out across developer platforms, enterprise tools like Vertex AI, and consumer applications like NotebookLM.

Google DeepMind is rolling out Lyria 3, a new generative music model, in beta within the Gemini app. The tool allows users to create high-quality, custom music tracks with lyrics and vocals based on text prompts or uploaded images.

Google DeepMind is launching a National Partnership for AI in India to collaborate with government bodies and local institutions. The initiative aims to deploy frontier AI capabilities to support national priorities in science, education, and public services.

Google DeepMind has released an upgraded version of Gemini 3 Deep Think, a specialized reasoning mode designed for complex scientific and engineering challenges. The update is available to Gemini AI Ultra subscribers and via API for select researchers and enterprises.

Google DeepMind has introduced Gemini Deep Think, a specialized mode designed to solve complex mathematical, scientific, and engineering problems. The model has demonstrated gold-medal standard performance in mathematics and programming competitions, moving beyond standard IMO problems to tackle professional-level research.

Google DeepMind is rolling out access to Project Genie, an experimental research prototype that allows users to create and interact with generative, physics-simulating worlds. The model functions as a general-purpose world model capable of real-time environment generation and interaction.

Google DeepMind has introduced D4RT, a new AI model designed for dynamic 4D scene reconstruction and tracking. The model aims to help machines understand 3D volumetric environments by processing 2D video sequences across both space and time.

Google DeepMind is updating its Veo model to improve consistency, creativity, and control when generating video from reference images. The update introduces native vertical outputs and high-fidelity upscaling up to 4K resolution for professional and mobile-first workflows.

Google reviews its major research and model breakthroughs for 2025, highlighting the evolution of AI from a tool to a utility. The report emphasizes the release of the Gemini 3 and Gemini 3 Flash models, which achieved significant advancements in reasoning and multimodal understanding.

Google has announced the release of Gemini 3 Flash, a new model designed for high speed, low latency, and cost-efficiency. The model is optimized for agentic workflows and is being rolled out across Google's developer platforms, consumer apps, and enterprise tools.

Google DeepMind has released Gemma Scope 2, an open-source suite of interpretability tools designed to help researchers understand the internal decision-making processes of Gemma 3 models. The toolkit aims to provide visibility into model behaviors to help debug issues like hallucinations, jailbreaks, and sycophancy.

Google has updated Gemini 2.5 Flash Native Audio to enhance live voice interactions and complex workflows. The update introduces improved natural conversation capabilities and live speech-to-speech translation features across Google products.

Google DeepMind is expanding its partnership with the UK AI Security Institute through a new Memorandum of Understanding. The collaboration focuses on foundational security and safety research to better evaluate and mitigate risks in advanced AI models.

Google DeepMind is deepening its partnership with the UK government to integrate frontier AI into science, education, and public services. The collaboration focuses on providing access to specialized models like AlphaEvolve and AlphaGenome to drive scientific discovery and national resilience.

Google DeepMind and Kaggle have introduced the FACTS Benchmark Suite to evaluate the factuality of large language models. The suite includes benchmarks for parametric knowledge, search tool usage, and multimodal accuracy to better measure how models handle information retrieval.

Researchers are using Google DeepMind's AlphaFold to predict the 3D structure of the GLYK enzyme. This research aims to engineer more heat-tolerant crops capable of surviving rising global temperatures and droughts.

Google DeepMind highlights the significant scientific impact of AlphaFold, noting its role in accelerating research in fields like plant physiology and drug discovery. The report emphasizes how the model has influenced tens of thousands of papers and fueled the development of new AI-driven drug design through Isomorphic Labs.

Researchers used Google DeepMind's AlphaFold to map the structure of the apoB100 protein, which is a key component of 'bad cholesterol'. This breakthrough provides critical insights into how LDL is formed, potentially transforming the treatment of cardiovascular disease.

Google DeepMind is supporting the U.S. Department of Energy's Genesis Mission, a national initiative to accelerate scientific discovery through AI. The mission aims to integrate AI with the DOE's 17 National Laboratories to solve complex challenges in energy and national security.

Google is integrating SynthID technology into the Gemini app to allow users to verify if images were created or edited using Google AI. Users can now upload images to Gemini and ask if the content is AI-generated to detect imperceptible digital watermarks.

Google has released Nano Banana Pro (Gemini 3 Pro Image), a high-fidelity image generation model designed for developers. The model features improved text rendering, world knowledge, and integration capabilities for platforms like Google AI Studio, Vertex AI, Adobe, and Figma.

Google DeepMind has introduced Nano Banana Pro, a new state-of-the-art image generation and editing model built on Gemini 3 Pro. The model features advanced reasoning and real-world knowledge to create accurate visuals, infographics, and diagrams.

Google DeepMind has introduced Gemini 3 Pro, a new highly intelligent model designed for advanced reasoning and agentic coding workflows. The model is available via the Gemini API in Google AI Studio and Vertex AI, offering improved performance in complex zero-shot tasks and terminal-based tool use.

Google DeepMind is opening a new research lab in Singapore to expand its presence in the Asia-Pacific region. The lab will focus on advancing Gemini's capabilities and developing AI solutions that are linguistically and culturally inclusive for the region.

Google DeepMind has introduced a new development or feature titled Google Antigravity. The specific details of the announcement are not provided in the excerpt.

Google DeepMind and Google Research have introduced WeatherNext 2, a highly efficient weather forecasting model capable of generating forecasts 8x faster with high resolution. The model is being integrated into Google products like Search, Gemini, and Google Maps, and is also available via Vertex AI and BigQuery.

Google DeepMind has introduced SIMA 2, an advanced AI agent capable of reasoning and interacting within 3D virtual environments. By integrating Gemini models, the agent evolves from a simple instruction-follower into a conversational companion that can set goals and learn through interaction.

Google DeepMind researchers have published a paper in Nature regarding how AI models organize visual representations compared to humans. The research introduces a method to align AI visual understanding with human knowledge to improve model robustness and generalization.

A six-month pilot program in Northern Ireland demonstrated that using Gemini and other generative AI tools saved teachers an average of 10 hours per week. The initiative focused on integrating AI to reduce administrative burdens in the education sector.

Google DeepMind is introducing new AI-driven research and tools to monitor and protect the Earth's biosphere. These include models for predicting deforestation risks, mapping species ranges, and advancing bioacoustic monitoring through the Perch model.