DeepMind

Latest AI news, models and releases from DeepMind. ['AlphaEvolve', 'AlphaFold', 'AlphaFold2', 'AlphaFold3', 'AlphaGenome', 'AMIE', 'Co-Scientist', 'DiffusionGemma', 'Gemini', 'Gemini 2.5 Flash', 'Gemini 3.5 Flash', 'Gemini 3.5 Flash Cyber', 'Gemini 3.5 Flash-Lite', 'Gemini 3.5 Flash-Lite 3.1', 'Gemini 3.5 Pro', 'Gemini 3.6 Flash', 'Gemini 3 Flash', 'Gemini Omni', 'Gemini Omni Flash', 'Gemini Spark', 'Gemma', 'Gemma 4', 'Gemma 4 12B', 'ReST^EM', 'SynthID', 'WeatherNext']

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 🇺🇸

Specialization Is Inevitable: What Optimization Theory, Biology, Markets, and Machine Learning Say

The article argues that specialization of AI systems is not simply a preference but an inevitable consequence of limited resources and performance demands. Arguments draw on the no-free-lunch theorem in optimization, evolutionary biology, competitive markets, and machine learning practice, including examples such as AlphaFold and Mixture-of-Experts.

DeepMind

Hugging Face blog27.07 · 09:05

Show more ↓