Google AI (Gemini)

Google’s Gemini models. Strong on multimodal (vision, audio), long context, and code.

Google AI’s Gemini models. Multimodal-first design — image, audio, video input alongside text.

Full walkthrough being written.

When to pick

• You need vision capabilities (image analysis, OCR).
• Audio input is a primary feature.
• You want very long context (1M+ tokens).

Setup

GOOGLE_AI_API_KEY in vly secrets. Use @google/generative-ai SDK from a Convex action.

Models

Model	Use for
gemini-2.5-pro	Maximum capability
gemini-2.5-flash	Fast, cheap

Related

OpenAI

Comparison.

Anthropic

Comparison.

Gemini CLI agent

The vly agent powered by Gemini.