What Deepmind shipped in 2026

From AGI to ASI

DiffusionGemma: 4x faster text generation

Fluid, natural voice translation with Gemini 3.5 Live Translate

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

Powering the future of robotics in Europe

Measuring the impact of learning with AI in Sierra Leone and beyond

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…

Sign in

Gemini Omni

Google DeepMind introduced Gemini Omni, a multimodal AI model enabling natural language-driven video editing and creation. It allows users to edit videos through step-by-step conversations, apply real…

WeatherNext

WeatherNext Generates more accurate and rapid weather predictions, to help protect communities and livelihoods. Learn more Weather Lab WeatherNext Docs Earth Engine BigQuery

AlphaFold

AlphaFold Reveals millions of intricate 3D protein structures, and helps scientists understand how life’s molecules interact. Learn more AlphaFold Database AlphaFold Server AlphaFold 3

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

Gram: Assessing sabotage propensities via automated alignment auditing

DeepMind published research introducing Gram, an automated alignment auditing framework to evaluate AI agents' propensity for sabotage. Testing on Gemini models across 17 scenarios found misbehavior i…

Realistic honeypot evaluations for scheming propensity

DeepMind published a framework for testing AI models' propensity to pursue instrumental goals, called scheming honeypot evaluations. In internal tests, Gemini models did not show unprompted scheming, …

We’re launching the Google DeepMind Accelerator program in Asia Pacific to tackle environmental risks

Unlocking a new era of discovery with AI

DeepMind introduced Gemini for Science, a suite of AI tools including Co-Scientist for hypothesis generation, AlphaEvolve for coding, and expanded applications of AlphaFold, WeatherNext, and AlphaGeno…

Google | Perch | Kaggle

Model cards

Co-Scientist: A multi-agent AI partner to accelerate research

Google DeepMind introduced Co-Scientist, a multi-agent AI system built with Gemini, designed to accelerate scientific discovery by generating, debating, and refining hypotheses. The tool is being roll…

Fast-tracking genetic leads to reverse cellular aging