Saturday, October 25, 2025

Start of archive·May 16

Top story

The Agent RaceGoogle DeepMind Blog

Gemini 2.5 Flash-Lite is now ready for scaled production use

Google's cost-efficient model moves from preview to production, expanding access to enterprise-grade capabilities at lower inference costs—a direct competitive move against OpenAI's smaller model tiers and Claude's pricing strategy.

Gemini 2.5 Flash-Lite now generally available (GA)

The briefs

Google expands its open-source model lineup with Gemma 3n, targeting developers with a focus on accessibility and community-driven iteration. This positions Google to compete directly in the open-source reasoning and inference space where Meta and Mistral dominate.

Gemma 3n announced by Google DeepMind

Google is making specialized, open-source AI models for healthcare, lowering barriers to entry for health-tech builders and positioning itself as infrastructure provider in a high-stakes vertical. This moves the model wars into domain-specific territory.

MedGemma collection announced

Google DeepMind releases T5Gemma, a new family of open-source encoder-decoder models that expand the Gemma lineup beyond decoder-only architectures. This matters for builders seeking alternatives to closed models and for companies optimizing for tasks like summarization, translation, and structured generation where encoder-decoder designs excel.

T5Gemma is part of Google's open-source Gemma model family

Google DeepMind's Veo video model moved from research artifact to production tool in a major filmmaker's workflow, signaling that generative video is crossing into professional creative infrastructure.

Google DeepMind partnered with director Darren Aronofsky on feature film 'Ancestra'
Enterprise AI news — Saturday, October 25, 2025 | KeyNews.AI