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 2.5', 'Gemini 2.5 Flash', 'Gemini-2.5 Flash', 'Gemini-2.5-Flash', 'gemini-2.5-flash-image', 'Gemini 2.5 Pro', 'Gemini-2.5 Pro']

Official site ↗

Regulation 🇷🇺

Jensen Huang's Coalition for Open Weights in AI Doubles: Google Joins, Amazon and Anthropic Stay Silent

Jensen Huang's letter "Open Weights and American AI Leadership" gathered 50 signatories in a day, including Google. Amazon and Anthropic have not joined yet; an Anthropic employee expressed skepticism on social media. Also discussed is the possible departure of Andrej Karpathy from Anthropic, which he denied.

Google/DeepMind

DeepMind

Anthropic

OpenAI

Amazon/AWS

NVIDIA

Habr — хаб ИИ26.07 · 16:02

Models 🇷🇺

Gemini Flash 3.6: Not Smarter, But Noticeably Cheaper

Google released three models: Gemini 3.6 Flash (main), Gemini 3.5 Flash-Lite (fast/cheap), and Gemini 3.5 Flash Cyber (security, restricted). The key improvement is a 17% reduction in output tokens for the same tasks, translating to 31-71% cost savings in agentic scenarios. However, the model's visual parsing regressed, especially on charts. Gemini 3.5 Flash-Lite is a surprise value leader, sometimes outperforming older mid-tier models. Google also announced internal testing of Gemini 3.5 Pro and the start of Gemini 4 pre-training.

Google/DeepMind

Habr — хаб ИИ26.07 · 15:02

Show more ↓