Sakhanda Wire
NVDA $153.77 -1.52% MSFT $504.91 -2.53% GOOGL $281.52 -0.78% META $620.99 -0.87% AMZN $256.80 -0.37%

DeepMind

29 stories · all sources

DeepMind

DeepMind launches EmbeddingGemma 2 for on-device multimodal search

Google DeepMind has introduced EmbeddingGemma 2, an open multimodal embedding model designed to run on local devices. Built on the Gemma 4 architecture and released under the Apache 2.0 license, it maps text, code, images, video and audio into a shared embedding space. The model has 740 million parameters.

DeepMind

Google DeepMind announces Gemini 4 Argon frontier model

Google DeepMind has announced Gemini 4 Argon, a frontier model designed for extended reasoning across software engineering, enterprise work and cybersecurity. The model is initially being provided to trusted cyber defenders through the Fairwind Program, while the company continues a phased safety process and gathers feedback before wider access.

DeepMind

SynthID Bio introduces watermarking for AI-designed proteins

A new proof of concept called SynthID Bio embeds an imperceptible, verifiable signature into AI-generated biological designs. The watermark is intended to remain detectable not only in a digital model, but also in a synthesized protein, while preserving biological function in laboratory tests.

DeepMind

Gemini 3.8 Live adds real-time visual avatars

Gemini 3.8 Live with Live Avatar combines the model’s live dialogue capabilities with near-real-time video generation. The feature gives enterprise conversational agents a visual persona that can listen, process what it sees and hears, and respond with synchronized speech and video. It is designed to support natural expressions, precise lip-syncing and fluid turn-taking.

DeepMind

Private AI Compute adds secure server-side memory

Google has outlined an update to Private AI Compute that is intended to provide persistent AI memory across devices while maintaining privacy standards associated with on-device processing. The company says the system will store information in dedicated, encrypted cloud storage, while the cryptographic keys needed to access it remain exclusively on a user’s personal devices.

DeepMind

Google introduces Gemini 3.8 Flash text-to-speech models

Google has introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, describing them as its most expressive audio-generation models so far. The models are designed for custom voice creation, scripted dialogue and multilingual audio experiences across Google AI Studio, the Gemini API, Gemini Notebook, Google Vids and, later, Gemini Enterprise.

DeepMind

DeepMind introduces Gemini 3.8 Live voice models

DeepMind has introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, two models designed for near-real-time voice interaction. The company says the first combines conversational capability, visual grounding and cost efficiency, while the Extended Thinking version targets complex tasks through deeper, multi-step reasoning.

DeepMind

AlphaGenome Atlas maps effects of 9 billion human DNA variants

Google DeepMind has introduced AlphaGenome Atlas, a platform containing predictions for the molecular effects of 9 billion possible single-nucleotide variants in the human genome. The resource is intended to address the difficulty of experimentally testing every possible single-letter DNA change and is available for academic research through a free website portal.

DeepMind

Google introduces WeatherNext 3 for higher-resolution global forecasts

Google DeepMind and Google Research have introduced WeatherNext 3, a global AI weather model that uses live geostationary satellite mosaics and updates forecasts every hour. The companies describe it as their most advanced and accurate global model so far, citing independent live evaluations by Brightband. It produces forecasts at several scales, including 5-kilometer detail for temperature and moisture, 10 kilometers for other surface variables and 25 kilometers for atmospheric variables. Overall, the system is about five times sharper than WeatherNext 2, which operated on a 25-kilometer grid every six hours.

DeepMind

Gemini 3.8 Flash models target reasoning, coding and cybersecurity

Google DeepMind has introduced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, describing them as its latest models for agentic workflows and cybersecurity. Gemini 3.8 Flash is presented as a general-purpose model for software engineering, autonomous agents and complex, multi-step reasoning. Its introductory price remains $0.75 per million input tokens and $3.75 per million output tokens, matching Gemini 3.7 Flash.

DeepMind

Google DeepMind launches agentic video understanding for Gemini

Google DeepMind has introduced agentic video understanding for Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite. The capability is designed to make video analysis more accurate while using fewer tokens and reducing costs. In the company’s benchmarks, it cut token consumption by up to 88%, lowered analysis costs by up to 66% and improved accuracy by up to 7%.

DeepMind

Gemini Omni 1.1 Flash adds controls for generative video

Google DeepMind introduced Gemini Omni 1.1 Flash, a set of generative-video capabilities and creative controls aimed at developers. The model is available through the Gemini API in Google AI Studio and is intended for video workflows, creative applications and media-editing software.