Who develops Gemini?

Gemini is Google's model family, developed by Google DeepMind, the research unit behind AlphaFold and Veo. Keying that from the name is easy. The part worth carrying away is which Google surface a developer actually reaches for, because choosing wrong burns an afternoon.

Four surfaces share the name. The Gemini app is consumer chat. Google AI Studio is the browser workspace where a key gets issued and prompts get prototyped; it creates a Cloud project and key automatically for new users. The Gemini Developer API is what application code calls with that key. The enterprise path, now called the Gemini Enterprise Agent Platform and previously the Vertex AI generative path, takes a Cloud project and a region instead, with its own auth and billing. Google's own guidance sends most developers to the Developer API unless specific enterprise controls are required.

OpenAI, Anthropic and Meta each ship a separate family behind separate endpoints and separate credentials. A Gemini key authenticates against none of them, and no SDK setting bridges that. Changing vendor means changing the endpoint and the account under it.