Google Gemini

Every agent runs on a model — the part that reads the request, works out what to do, and writes the answer back. Pluto serves multiple Google Gemini models, and you choose one per build in Agent Studio under Agent Model.

Whichever you pick, everything around it stays the same:

• The same Response-Agent API call, with the same fields.
• The same tools your agent is configured with.
• The same purpose, instructions, and guardrails from your build.

What changes is how capable your agent is, and what your API calls cost.

Supported Models

Capability Explained

Capability is a model’s core processing power—specifically, how much information it can connect and manage at one time. Think of it as the model's capacity for complex logic.

A highly capable model excels at:

• Complex Instructions: Following long, nuanced, or strict guidelines without losing track of the goal.
• Multi-Step Workflows: Planning and executing tasks that require several consecutive steps to solve.
• Ambiguity: Making sense of user inputs that are vague, disorganized, or oddly worded.
• Tool Usage: Deciding when to use an external tool, formatting the correct request for it, and accurately interpreting the data that comes back.

Data Policy

The data processed by your AI agent will never be used to train AI models.

The prompts, uploaded files, session history and the model's generated responses will never be included in datasets used to train or improve AI models.

You retain full ownership of all your inputs as well as the outputs generated by your agent.