Vertex AI

Vertex AI is the former name of Gemini Enterprise Agent Platform; the rename kept the endpoint aiplatform.googleapis.com in place. Its defining property across the source notes is multi-model selection. Alan Blount, Technical Solutions Consultant at Google Cloud, showed on 14 September 2026 that Claude Fable 5.1 and Gemini 3.8 Flash sit behind the same API, reached through Model Garden. Activation is not symmetric between them. Fable falls under Google Cloud's Advanced AI Safety Addendum: without prompt-response sharing enabled for Anthropic at project level through a setPublisherModelConfig call, every request returns a 403. Flash works as soon as the model is turned on.

The development-versus-production split recurs. The google-agentic-commerce/AP2 repository documents two authentication paths: a Google API key, recommended for development for its simplicity, and a Vertex AI configuration advised for production deployments. Gemini 2.5 Flash-Lite reached stable general availability on 22 July 2025 in Google AI Studio and Vertex AI.

The counterweight is where the notes keep landing. Janakiram MSV, writing on 20 July 2026, places Gemini Enterprise Agent Platform beside Bedrock AgentCore and Microsoft Foundry as three convergent enterprise agent architectures with no portability contract: session state, traces and identity all terminate with a single provider, and moving an agent a year later means rebuilding everything. Mathieu Grymonprez, Global CDO of Adeo, names the lock-in of enterprise intelligence as his largest concern. The platform advised for production is also where that question concentrates.