Google Cloud

On 8 October 2026 at Gemini at Work 2026, Thomas Kurian presented the Gemini agent and split apart two things usually sold together: the agent and the model running inside it. Gemini chooses per task between Gemini models and Anthropic's Claude models, with others to follow. Google Cloud's stated reasoning is that the best model changes every few months.

Alan Blount, Technical Solutions Consultant at Google Cloud, worked the same idea at implementation level in a four-step tutorial on @GoogleCloudTech: Claude Fable 5.1 and Gemini 3.8 Flash behind the single API of Gemini Enterprise Agent Platform, the new name for Vertex AI. Fable falls under the Advanced AI Safety Addendum, so without project-level prompt-response sharing enabled for Anthropic, every request returns a 403.

Google Cloud also supplies much of the evidence base on returns. It publishes the Rapport DORA 2025 and, in April 2026, a 60-page ROI framework built on the claim that AI amplifies what an organisation already does, well or badly. The numbers it carries: 727% average three-year ROI for Google Cloud AI customers, 78% of executives reporting ROI on at least one gen AI use case, 88% of agentic AI early adopters seeing positive returns, a 98% fall in price per token since 2024. The same document advises against headcount-reduction strategies and tells readers to budget for a J-curve trough first.

S3NS, the Thales joint venture qualified SecNumCloud 3.2 in December 2025, carries that stack to regulated French buyers: AG2R LA MONDIALE's Almia platform runs on it.