Do Son Google revealed Gemini 4 Carbon, a flagship model dominating benchmarks with zero prompt injection flaws and autonomous cybersecurity capabilities.
Google DeepMind
Google DeepMind skips v4.0 to drop Gemini 3.5 Flash. Running 4x faster than rivals, it powers 24/7 autonomous agents via the new Antigravity framework.
Google has cut the Gemini API Tier 2 spend requirement to $100 and added per-project spend caps. Learn how to upgrade faster and prevent "billing catastrophes."
Google DeepMind’s Gemini Embedding 2 is here. Map text, images, 120s of video, and native audio into one vector space to revolutionize multimodal RAG and search.
Boston Dynamics and Google DeepMind are bringing Gemini AI to the new Atlas robot. Hyundai plans to deploy these "intelligent" humanoids by 2028.
Samsung's 2026 TVs will natively integrate Google Photos and Vision AI, bringing exclusive Memories, AI editing, and video generation to the big screen.
YouTube launches the Playables Builder beta: an AI tool powered by Gemini 3 that lets creators build mini-games using only natural language prompts.
DeepMind is partnering with CFS to optimize its SPARC fusion reactor for net energy gain (Q>1) by 2027. The AI uses TORAX simulations and AlphaEvolve agents to control extreme plasma heat.
Google DeepMind's Dustin Tran, a core developer of the Gemini project, has left the company to join Elon Musk's xAI to work on the next generation of Grok.
OpenAI postponed its open-weight AI model release due to security concerns. Separately, its WindSurf acquisition reportedly collapsed, with top executives joining Google DeepMind for $2.4B.
Isomorphic Labs (Google DeepMind's subsidiary) announced its first AI-designed drug is ready for clinical trials, marking a historic milestone in AI-driven pharmaceutical development.
Google announces Gemini AI support for Anthropic's Model Context Protocol (MCP), aiming to standardize AI model integration with external tools and data, enhancing interoperability.