DeepMind 在 2026年4月 发布的内容

Gemini Robotics

DeepMind introduced Gemini Robotics, a suite of models including the vision-language-action (VLA) model 1.5 and the embodied reasoning (ER) model 1.6, designed to enable robots to perceive, reason, an…

Gemini Live API overview

Google introduced the Gemini Live API, enabling low-latency, real-time voice and vision interactions with its Gemini model. The API supports continuous audio, image, and text streams to deliver immedi…

Vertex AI Studio

Vertex AI Studio Explore 200+ models on our enterprise platform with tools and features for AI development Try Vertex AI Studio

Google Antigravity

Google Antigravity Experience an agentic development platform, evolving the IDE into the agent-first era Try Google Antigravity

Gemini Audio

New Gemini Audio Advanced real-time audio models, built on Gemini Try it in Google AI Studio Vertex AI Studio Learn more

Nano Banana 2 🍌

Nano Banana 2 🍌 Powerful image generation, advanced intelligence, and enhanced creative precision – with the speed you expect from Flash. Learn more Try

Announcing our partnership with the Republic of Korea

Google DeepMind and Korea partner to accelerate scientific breakthroughs using frontier AI models

Dynamic Reflections: Probing Video Representations with Text Alignment

The study introduces the first comprehensive examination of video-text representation alignment, revealing that cross-modal alignment is highly dependent on the richness of both visual and text data. …

State-of-the-art multimodal embedding model

Google DeepMind introduced Gemini Embedding 2, a state-of-the-art multimodal embedding model that maps text, images, videos, audio, and documents into a unified embedding space for semantic understand…

A tool to watermark and identify content generated through AI

Google introduced SynthID, a watermarking tool for AI-generated content (images, audio, text, video) embedded directly into outputs from its generative AI products. The watermarks are imperceptible to…

Image Generators are Generalist Vision Learners

DeepMind’s research demonstrates that image generation training enables models to develop strong visual understanding, rivaling specialized vision models like Segment Anything and Depth Anything. Visi…

Decoupled DiLoCo: A new frontier for resilient, distributed AI training

Decoupled DiLoCo: A new frontier for resilient, distributed AI training

Partnering with industry leaders to accelerate AI transformation

Google DeepMind announced a partnership with global consultancies to accelerate AI adoption for organizations worldwide. The collaboration aims to integrate frontier AI technologies into enterprise wo…

Gemini Embedding 2Stay organized with collectionsSave and categorize content based on your preferences.

Google released Gemini Embedding 2, a multimodal embedding model generating 3072-dimensional vectors from text, images, documents, audio, and video inputs. It introduces custom task instructions for o…

Gemini 3.1 Flash TTS: the next generation of expressive AI speech

DeepMind has introduced Gemini 3.1 Flash TTS, a next-generation AI speech model that enables granular audio tags for precise control over expressive audio generation. This advancement allows users to …

Our most intelligent open models, built from Gemini 3 research and technology to maximize intelligence-per-parameter

Google DeepMind introduced Gemma 4, its most intelligent open models derived from Gemini 3 research, optimized for intelligence-per-parameter efficiency. The models are designed for maximum compute an…

Gemini Robotics-ER 1.6: Powering real-world robotics tasks through enhanced embodied reasoning

DeepMind announced the release of Gemini Robotics ER 1.6, a significant update to its robotics model focused on improving spatial reasoning and multi-view understanding for autonomous robots. This ver…

Generate music with LyriaStay organized with collectionsSave and categorize content based on your preferences.

Google Cloud introduced Lyria, a new text-to-music generation model on Vertex AI, enabling users to create novel music tracks from text prompts via the Google Cloud console or Gemini API. The feature …

Google modelsStay organized with collectionsSave and categorize content based on your preferences.

Google Cloud’s Vertex AI has expanded its generative AI model lineup with new and updated models, including the high-performance 3.1 Pro for complex problem-solving, the agentic and coding-focused 3 F…

Gemma 4: Byte for byte, the most capable open models

DeepMind has introduced Gemma 4, its most advanced open-source AI models to date, designed for enhanced reasoning capabilities and agentic workflows. The new models represent a significant leap in per…

Google introduced Gemma 4, its most advanced open AI model family to date, designed for advanced reasoning and agentic workflows. The release includes four model sizes—Effective 2B (E2B), Effective 4B