Google/DeepMind

Latest AI news, models and releases from Google/DeepMind. ['A2UI v0.9', 'AI Co-Scientist', 'AI Evaluator', 'AI Mode', 'AI Overviews', 'Alexa', 'AlphaEvolve', 'AlphaFold', 'AlphaGenome', 'AlphaQubit', 'Antigravity', 'Auto frame', 'BERT', 'BioBERT', 'Chinchilla', 'ClinicalBERT', 'Computational Discovery', 'Confidential GKE Nodes', 'Co-Scientist', 'Empirical Research Assistance', 'Era', 'ERA', 'Executive LLM', 'Farmscapes 2020', 'Flood Hub', 'frontier AI', 'Frozen v2', 'FunctionGemma', 'Gemini', 'Gemini 1.5 Flash', 'Gemini-1.5-pro', 'Gemini 1.5 Pro', 'Gemini 2.0 Flash', 'Gemini 2.0 Flash-Lite', 'Gemini 2.5', 'Gemini 2.5 Flash', 'Gemini-2.5 Flash', 'Gemini-2.5-Flash', 'gemini-2.5-flash-image', 'gemini-2.5-flash-lite']

Official site ↗

Models 🇺🇸

DiffusionGemma: 4x faster text generation

Google DeepMind has introduced DiffusionGemma, an experimental open 26B Mixture of Experts model that uses text diffusion to generate up to 4x faster than autoregressive LLMs on GPUs. It simultaneously processes 256-token blocks, targeting speed-critical local workflows like in-line editing, and is released under Apache 2.0 on Hugging Face.

Google/DeepMind

DeepMind

NVIDIA

Google DeepMind27.07 · 12:02

Research 🇺🇸

New Framework for Auditing Machine Unlearning

Google researchers introduced Regularized f-Divergence Kernel Tests, a new framework for auditing machine unlearning that allows statistically robust verification of whether a model has forgotten confidential data. The method addresses issues with traditional tests, such as loss of sensitivity on large models and false positives. The framework has been successfully tested on synthetic data, high-energy physics tasks, differential privacy auditing, and unlearning evaluation.

Google/DeepMind

Google Research27.07 · 11:06

Show more ↓