Ce que DeepMind a publié en 2026

Gemini Robotics 2

Your browser does not support the video tag. Your browser does not support the video tag. Gemini Robotics 2 The intelligence layer to power any kind of robot Learn more Try

Gemini Robotics 2 brings whole body intelligence to robots

Gemini Robotics 2 brings whole body intelligence to robots

Visual prompt engineering for video models

DeepMind’s research proposes visual prompt engineering (VIPE), a technique to automatically modify task images to improve video model performance. Testing shows VIPE outperforms text-based prompt engi…

Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

We’re introducing new Gemini models, including Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber.

Towards Structural Understanding of LLM Overthinking

DeepMind published research introducing TRACE, a fine-grained analyzer that decomposes LLM thought processes into sub-thoughts and maps discourse relationships to identify overthinking patterns. The s…

Gemini Enterprise Agent Platform

Gemini Enterprise Agent Platform Build, scale, and govern agents Try Gemini Enterprise Agent Platform

Nano Banana 2 Lite 🍌

Nano Banana 2 Lite 🍌 Our fastest, most efficient Gemini Image model Learn more Try

Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration

Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration July 2026 Models Learn more

Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber July 2026 Models Learn more

We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control

We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control July 2026 Models Learn more

DeepMind introduced Nano Banana 2 Lite, a faster and more cost-efficient image generation and editing model under the Gemini Image suite. It delivers high-speed generation with lower latency and reduc

Bridging the Scale Gap: Augmenting Human Red-Teaming to Uncover Latent Risks in T2I Models

DeepMind published research introducing Seed2Harvest, a hybrid framework that combines human red-teaming with automated prompt expansion to uncover latent risks in Text-to-Image (T2I) models. The appr…

Securing the future of AI agents

Securing internal systems with an AI Control Roadmap, combining traditional safeguards and real-time monitoring.

Investing in multi-agent AI safety research

Google DeepMind and partners announce a $10M funding call for multi-agent safety research.

Solipsistic superintelligence is unlikely to be cooperative

DeepMind’s research contends that traditional AI design, which treats the world as a static feedback source, risks producing superintelligent systems that are non-cooperative. The paper introduces the…

Introducing Gemma 4 12B: a unified, encoder-free multimodal model

Introducing Gemma 4 12B: a unified, encoder-free multimodal model June 2026 Models Learn more

DiffusionGemma: 4x faster text generation

DiffusionGemma: 4x faster text generation June 2026 Models Learn more

Introducing computer use in Gemini 3.5 Flash

Introducing computer use in Gemini 3.5 Flash June 2026 Models Learn more

Fluid, natural voice translation with Gemini 3.5 Live Translate

Fluid, natural voice translation with Gemini 3.5 Live Translate June 2026 Models Learn more

Introducing Gemini for Science

Your browser does not support the video tag. Your browser does not support the video tag. Introducing Gemini for Science Learn more