The latest AI news and updates of Google — the company behind Gemini, DeepMind research, Gemma open models, and AI across Search, Cloud, and Workspace. Covering Google's latest product updates, launches, and research from the past 90 days.
Google14h agoGoogle released Gemini 3.5 Live Translate, an audio model for near real-time speech-to-speech translation across 70+ languages. The model streams continuous audio output with synced transcripts, preserving speaker intonation and pacing. Developers can access the model in public preview via the Gemini Live API and Google AI Studio, while consumer availability rolls out to the Google Translate app.
Gemini16h agoGoogle expanded Notebooks in Gemini to the European Economic Area, United Kingdom, and Switzerland. The feature provides a dedicated space for projects, allowing users to organize sources, instructions, and chat history in one place. Users can now create and manage these persistent project notebooks directly within the Gemini app or web interface.
Google DeepMind16h agoGoogle DeepMind is partnering with Brazilian football club SE Palmeiras to implement TacticAI, an AI system that predicts open play dynamics up to 8 seconds in advance. Using graph neural networks to map player interactions, the system allows the club’s data science team to simulate field scenarios and test defensive setups by dragging and dropping players in real time.
Google DeepMind16h agoGoogle DeepMind, Schmidt Sciences, the Cooperative AI Foundation, and ARIA launched a 10 million dollar research fund to study collective behaviors in large-scale multi-agent systems. Researchers can submit proposals for sandboxes, agent network science, infrastructure, and oversight by August 8, 2026. Awardees will be announced in Autumn 2026.
Google DeepMind16h agoGoogle DeepMind published results from an eight-week randomized controlled trial in Sierra Leone involving 1,763 students. The study found that Gemini’s Guided Learning feature shifted student behaviour, with queries about problem-solving methods rising from 68% to 90%. By using scaffolding questions instead of direct answers, the AI helped students achieve math score gains of 0.258 standard deviations.
Google Gemma16h agoGoogle released DiffusionGemma, an experimental open model that generates text using diffusion instead of sequential token prediction. By generating 256 tokens in parallel, it delivers up to 4x faster inference on dedicated GPUs, exceeding 1000 tokens per second on an H100. This 26B Mixture of Experts model supports real-time self-correction for tasks like code infilling and in-line editing.
GoogleJun 10Google AI released Gemini 3.5 Live Translate, an audio model for live speech-to-speech translation. It supports over 70 languages, streaming translations continuously to maintain natural conversation flow by preserving speaker intonation and pacing. This aims to eliminate awkward pauses, making cross-language interactions feel more fluid across various applications.
NotebookLMJun 9Google's NotebookLM is rolling out upgrades to Google AI Ultra subscribers, introducing agentic capabilities in chat, more advanced reasoning, and new output formats. These enhancements enable the tool to tackle complex, multi-step research problems more autonomously, offering deeper insights and personalized content generation.
GoogleJun 8Google Research and Google DeepMind's paper on D4RT, a model for dynamic 4D scene reconstruction, received the CVPR 2026 Best Paper Award. This recognition highlights a new feedforward approach that efficiently reconstructs complex geometry and motion from video, unifying multiple tasks with improved speed and accuracy.
GoogleJun 7Google Research introduced 3DCodeBench, a new benchmark evaluating AI models' ability to generate 3D objects using code. This benchmark, presented at CVPR2026, demonstrates how agentic AI can autonomously create complex 3D assets, highlighting the role of iterative refinement in improving model performance.
GoogleJun 7Google DeepMind is presenting TIPSv2, a new foundational image-text encoder, at CVPR 2026. This model enhances spatial awareness and patch-text alignment, improving performance across vision and multimodal applications, including strong gains in zero-shot segmentation.
GoogleJun 7Google Research and Google Cloud introduced a new agentic RAG framework designed to handle complex enterprise queries. This framework employs a multi-agent workflow that iteratively searches for sufficient context, improving accuracy beyond standard Retrieval-Augmented Generation (RAG). It aims to deliver more dependable responses by preventing the AI from guessing when information is incomplete across multiple data sources.
GoogleJun 7Google Research is introducing D4RT, a unified AI model that reconstructs and tracks dynamic scenes across space and time from a single video. This model advances computer vision by efficiently inferring depth, spatio-temporal correspondence, and camera parameters, setting a new state of the art for 4D scene understanding.
GoogleJun 7Google Research announced BlazeEdit, an efficient, generalist image-to-image diffusion model designed for on-device mobile deployment. This model enables interactive image editing tasks like outpainting and relighting directly on mobile devices, addressing the computational and privacy challenges of server-side AI.
NotebookLMJun 7Google's NotebookLM has launched a new Source Attribution feature. This update provides visibility into the exact prompts and source materials that generate AI outputs, enhancing transparency and control over the content. It also enables customization of these underlying "formulas" to refine AI-generated artifacts.
NotebookLMJun 7Google's NotebookLM introduced a new "Sherlock Holmes notebook" that transforms studying into an interactive mystery game. This feature allows users to engage with source material by deducing facts and solving a murder within a text adventure, shifting learning from passive reading to active investigation.
NotebookLMJun 7Google's NotebookLM mobile app now allows users to create briefing documents, study guides, and blog posts directly from their phones. This update extends AI-powered content generation capabilities to mobile, enabling users to produce structured reports and educational materials on-the-go.
GeminiJun 5Gemini Live now lets users create and edit images directly within the app, using a live camera feed. This brings AI into real-time visual interactions, turning spoken or typed instructions into immediate on-screen changes.
GoogleJun 5Google released new Gemma 4 Quantization-Aware Training (QAT) checkpoints, including GGUF (Q4_0) and a custom mobile schema under 1GB. These enable running Gemma 4 models locally on consumer GPUs and mobile devices with reduced memory footprint and accelerated decode speeds, while preserving reasoning quality.
GeminiJun 5The Gemini app is now available as a native macOS experience, allowing users to share their active window with the AI assistant via a keyboard shortcut. This enables context-aware help directly within workflows, eliminating the need for manual screenshots or switching tabs.
Google GemmaJun 4Google's Magenta Project released Magenta RealTime 2 (MRT2), an open-weight, open-source live music model. It enables low-latency, real-time music synthesis natively on Apple Silicon MacBooks using MIDI, text, and audio inputs. This allows musicians to play AI-generated music as an instrument directly on their device, fostering new creative workflows.
GoogleJun 4Google released Gemma 4 12B, a unified multimodal model that processes audio and vision directly within the LLM backbone. It brings near-frontier reasoning to consumer hardware, enabling complex agentic workflows to run entirely offline on standard laptops.
Google LabsJun 4Google Labs released Dreambeans, an experimental mobile app that uses a Personal Intelligence layer to synthesize data from across the Google ecosystem into daily visual stories. The app shifts AI from reactive chat to proactive surfacing, using a user's own photos and history to create personalized content.
Google GemmaJun 2Google launched the first iteration of gemma-skills, an open-source library of reusable capabilities for AI agents. By standardizing how agents select model sizes and use performance optimizations like MTP, Google is making it easier to build efficient, autonomous workflows on top of the Gemma ecosystem.
Google DeepMindJun 2Google DeepMind introduced Co-Scientist, a multi-agent system built on Gemini that generates and debates scientific hypotheses. The system moves beyond simple literature search by using a tournament of ideas to refine and rank novel research leads. Researchers can now access these capabilities through the new Hypothesis Generation tool.
Google AI StudioJun 2Google has launched native integrations for Gmail, Drive, and Sheets within the AI Studio interface. Users can now build and test applications that interact with live Workspace data using natural language prompts.
GeminiJun 1Google released a personal avatar feature for Gemini Omni that lets users create digital twins of their face and voice. By moving from generic generation to persistent personal likenesses, Google is turning its assistant into a personalized production studio where users can star in content via simple text prompts.
GeminiMay 29Google rolled out Gemini Omni Flash to AI subscribers and YouTube Shorts, enabling users to transform sketches and existing footage through natural language dialogue. The model uses multimodal reasoning to maintain physical consistency and character memory across multiple rounds of video edits.
GeminiMay 29Google began rolling out Gemini Spark, a 24/7 personal AI agent, to all Google AI Ultra subscribers in the U.S. The agent executes multi-step workflows autonomously in the background, maintaining persistence even when a user's physical devices are powered off.
Google GemmaMay 29Google released the Google AI Edge Gallery app and LiteRT-LM framework to enable fully offline agentic workflows on mobile and IoT devices. By running Gemma 4 locally, developers can build multi-step agents that plan, use tools, and process multimodal data without cloud latency or privacy risks.
GeminiMay 28Google rolled out Gemini Omni's video-to-video editing capabilities to users in India, allowing direct uploads from camera rolls for AI-powered transformations. This expansion brings high-fidelity video remixing to the region, allowing users to transform personal media directly within the assistant app.
GoogleMay 28Google moved its Nano Banana image generation models, including Nano Banana 2 and Nano Banana Pro, to General Availability via the Gemini API. This transition from preview to production-ready status enables developers to integrate high-fidelity 4K visuals and real-time search grounding into stable applications.
Google GemmaMay 28Google's Gemma 4 31B ranked as the top-performing open-weight model on TERMS-Bench, a new evaluation for AI agents conducting economic negotiations. The benchmark uses a verifiable environment instead of LLM grading to measure an agent's ability to maximize profit while following strict financial constraints.
Google GemmaMay 27Google and Synaptics announced the Coralboard, a development platform featuring a new integrated Neural Processing Unit for local AI acceleration. This shift to an open-source RISC-V architecture allows developers to run complex multimodal models entirely on-device without cloud dependency.
GeminiMay 27Google rolled out a new video composition feature for Gemini Omni that turns text, video, and up to five images into a single ten-second clip. This shift moves AI video from simple generation to active asset remixing directly within a general-purpose assistant.
Google DeepMindMay 23Google expanded its national AI partnership with Singapore to deploy frontier models across healthcare, education, and scientific research. The initiative shifts AI from a general-purpose tool to a sovereign infrastructure layer, aiming to add S$3.3 billion in economic value by 2040.
Google DeepMindMay 22Google is expanding its SynthID invisible watermarking technology to partners like OpenAI and ElevenLabs while launching native verification tools in Search and Chrome. These updates allow users to identify AI-generated media directly through conversational queries and verify authentic camera-captured content across platforms like Instagram.
Google GemmaMay 22Google demonstrated Gemma 4 E4B autonomously navigating an iOS simulator using the Argent framework. This shift proves that lightweight, open-weight models can handle complex software interactions locally, reducing the need for cloud-based computer use.
Google DeepMindMay 22Google DeepMind's Project Genie now allows users to transform real-world U.S. locations from Google Maps Street View into interactive, navigable 3D environments. By anchoring generative world models in real-world imagery, the update shifts AI simulation from purely imaginative landscapes to playable versions of actual places.
Google GemmaMay 22Google demonstrated an experimental Gemma 4 application running entirely offline on a Pixel phone and prototype display glasses. The field test proves that complex multimodal tasks like visual understanding and tool use can function without any cloud connectivity or data service.
GeminiMay 21Google expanded the Gemini app's ecosystem to include direct connections with OpenTable, Canva, and Instacart. The update shifts the assistant from information retrieval to task execution, allowing users to book reservations and design assets within the chat interface.
GoogleMay 21Google updated Stitch, its AI-native design canvas, to support live streaming builds and direct codebase imports for brand-consistent UI generation. The tool now integrates with the Model Context Protocol to enable autonomous collaboration between design layouts and external AI coding agents.
Google AI StudioMay 21Google expanded AI Studio into a full-scale development environment that builds native Android apps and integrates directly with Workspace data. By removing the need for local SDKs and high-performance hardware, Google is shortening the path from a natural language prompt to a production-ready mobile application.
GoogleMay 21Ramp used the new Managed Agents in the Gemini API to build and deploy advanced financial agents without managing backend infrastructure. This shift allows teams to offload complex agent orchestration and state management to Google, significantly reducing the engineering overhead required for production-grade autonomous workflows.
GeminiMay 21Google is rolling out Gemini 3.5 Flash globally to all users for free via the web and mobile app. The update shifts the high-speed model from a developer tool to a consumer assistant capable of analyzing complex diagrams and math papers. This move democratizes frontier-level multimodal reasoning for everyday research and document exploration.
Google AI StudioMay 21Gemini 3.5 Flash took the top spot on Zapier's Automation Bench, outperforming all other frontier models in operations and support tasks. The result validates Google's strategy of delivering high-speed, low-cost models that maintain competitive intelligence for autonomous workflows.
Google AI StudioMay 21Gemini 3.5 Flash has ranked first on the APEX-Agents-AA benchmark, outperforming larger frontier models in autonomous task execution. The result confirms that high-speed, low-cost models are now capable of handling complex agentic workflows previously reserved for larger architectures.
GoogleMay 21Google open-sourced its Science Skills toolkit on GitHub to provide AI agents with grounded scientific data and improved token efficiency. The release includes specialized capabilities for genomics and structural biology, integrating over 30 databases including AlphaGenome and UniProt.
Google AI StudioMay 20Google AI Studio now supports native connectors for Google Workspace, allowing apps to directly pull data from Sheets and manage files in Drive. This update transforms the prototyping environment into an operational hub where AI agents can interact with a user's live document ecosystem.
Google AI StudioMay 20Google AI Studio now supports building Android applications using natural language prompts and testing them directly on physical devices. This update moves AI-assisted development from web-based prototypes to native mobile software that can be validated on actual hardware.
GeminiMay 20Google released a new feature for Gemini Omni that allows users to create persistent AI digital twins of their own voice and likeness. By storing these avatars, users can generate personalized video content without re-uploading reference media for every session. This move brings high-fidelity video personalization directly into a general-purpose AI assistant for the first time.
Google AI StudioMay 20Google is launching a mobile version of AI Studio for iOS and Android to enable rapid application prototyping on the go. The app allows users to generate functional web tools from natural language prompts and deploy them to shareable URLs with a single tap.
GoogleMay 20Google launched a suite of AI-powered creative tools including Google Pics for Workspace and an autonomous agent for the Google Flow video platform. These updates shift AI from simple asset generation to multi-step project planning and natural language tool creation.
Google DeepMindMay 20Google updated its Flow creative studios with Gemini Omni Flash to enable precise video editing and stable character identities across scenes. By introducing an autonomous agent for batch editing and natural language tool creation, Google is shifting AI video from single-clip generation to a managed production workflow.
Google DeepMindMay 20Google DeepMind introduced Gemini for Science, a suite of experimental tools designed to assist researchers with literature analysis, hypothesis generation, and computational modeling. By moving beyond simple chat to multi-agent tournaments and autonomous code iteration, Google is verticalizing its frontier models for high-stakes scientific discovery.
GoogleMay 20Google launched a unified AI Search experience that merges AI Overviews and AI Mode into a single conversational interface powered by Gemini 3.5. The update enables users to query across text, images, files, and video while maintaining persistent context for follow-ups.
GoogleMay 20Google introduced managed agents for the Gemini API, allowing developers to deploy autonomous workflows through a single API call. By handling the underlying orchestration and infrastructure, Google is lowering the technical barrier for moving agentic prototypes into production environments.
GeminiMay 19Google released Daily Brief, a personalized morning digest that synthesizes data from your inbox, calendar, and tasks into a prioritized plan. The feature shifts Gemini from a reactive assistant to a proactive agent that suggests next steps before you start your workday.
Google DeepMindMay 19Google DeepMind introduced Gemini Omni Flash, a multimodal model that allows users to transform existing video scenes using natural language prompts. By combining generative media systems with Gemini's reasoning, the model can instantly swap environments or add objects while maintaining the original video's action.
Google DeepMindMay 19Google DeepMind expanded its Antigravity ecosystem with a new desktop orchestrator, a developer SDK, and API-managed agents that run in isolated Linux sandboxes. By pairing these tools with the high-speed Gemini 3.5 Flash model, Google is shifting AI development from single-turn chat to autonomous, multi-step agentic engineering.
Frequently asked questions