Multi-model (Claude + Gemini)

Overview

The landscape of AI models is always evolving, but there can consistently be a best model system—one that leverages multiple models to maximize value for clients. Harvey is a multi-model provider platform designed to deliver the best model for each use case, offering models from OpenAI, Anthropic (Claude model suite), and Google (Gemini model suite) to support a wide range of needs within the product.

Using Models and Options

For most users, this change will only be felt in results, and there is no action or model selection needed on their end—Harvey will automatically use the best model for the task at hand.

For advanced users who opt in, a model selector will be available in Assistant composer and Vault review table columns. Users can choose between these model options:

• Auto - Harvey will select the best model for the task (Recommended)
• GPT-4.1 - General purpose tasks
• o3 - Complex tasks, slower
• Claude 3.7 Sonnet - General-purpose tasks
• Gemini 2.5 Pro - Complex tasks, slower

Model Selector in Assistant

In Assistant, the desired model can be selected through the selector icon before clicking Ask Harvey and submitting your query.

Model Selector in Vault Review Table

Use the Model dropdown menu to select a different model when creating a new column query.

FAQ

If I select auto, will I be notified which model is being used?

No. Harvey often uses multiple models in one task, making it too complex to expose this clearly.

If I choose a specific model, does that mean only that model will be used to answer my query?

No, our model system will still use the full model suite that powers it today. Only the final answer model that is used will be changed with your model selection.