Von Logo
Documentation
DocumentationModel Selector

Model Selector

Model selector allows users to select models. We want to harbor transparency and also provide options for cheaper models and free models for users to be able to have more power.

Model Selector

Von includes a Model Selector feature that allows you to choose between different AI models for your tasks. Whether you're generating code, asking technical questions, or iterating on features, selecting the right model can improve both the speed and quality of your results.

Selecting AI Models

The Model Selector gives you control over which large language model (LLM) powers your conversation. Available models include:

  • Claude 3.7 Sonnet
  • Claude 3.7 Sonnet:thinking (deeper reasoning mode)
  • Claude 3.5 Sonnet
  • Gemini 2.5 Pro
  • Gemini 2.5 Flash:thinking (enhanced reasoning mode, Free)
  • GPT-4o (2024-11-20)
  • Von (Default)

Some models provide :thinking modes, designed for more deliberate reasoning and better handling of complex tasks like architecture planning or multi-step logic.


Credit System

Each model has an associated credit cost per message, clearly shown in the selector. Credits are only deducted when you send a message.

ModelCost per message
Claude 3.7 Sonnet10 credits
Claude 3.7 Sonnet:thinking10 credits
Claude 3.5 Sonnet10 credits
GPT-4o (2024-11-20)10 credits
Gemini 2.5 Pro Preview5 credits
Gemini 2.5 FlashFREE
Gemini 2.5 Flash:thinkingFREE
Von (Default, with smart routing)10 credits (may vary based on routing)

New models will be added as they become available, and we can add specific models on request. If you need a particular model not listed here, just ask!


Von as the Default Model

Von is the default selection — but it’s more than just one model. Von uses a smart routing system that analyzes:

  • Your prompt content
  • Project context and codebase
  • Complexity of the request
  • Task type (e.g., code generation, explanation, refactoring, architecture design)

Von dynamically selects the best underlying model (or combination of models) for the situation. This means you get the power of multiple models without needing to manually choose every time.

Think of Von as your AI teammate who knows when to bring in the heavyweight models and when faster or more cost-effective options will do.

If you're unsure which model to pick, use Von — it handles intelligent model selection for you.


Use Cases for Manual Selection

While Von’s smart routing handles most situations automatically, there are cases where manual selection may be helpful:

✅ Deep Reasoning and Architecture Planning

For thoughtful, multi-step reasoning — like breaking down architectures, exploring trade-offs, or complex algorithm design — directly selecting Claude 3.7 Sonnet:thinking or Gemini 2.5 Pro can give better results.

✅ Quick, Low-Cost Experimentation

For brainstorming, drafts, or quick testing where you want to conserve credits, Gemini 2.5 Flash (FREE) or Gemini 2.5 Flash:thinking (FREE) are excellent options.

✅ Comparing Models for Quality and Behavior

If you want to compare how different models approach the same problem, manually switching between models allows you to evaluate their strengths and trade-offs directly.

✅ When You Have a Favorite Model for Certain Tasks

If a specific model consistently delivers the best results for your frontend, backend, or infrastructure tasks, manual selection gives you that control.


How to Use the Model Selector

  1. Click on the model dropdown in the message input box of your Von workspace.
  2. Select your preferred model for the current session.
  3. Credits will be deducted per message based on the model.
  4. Switch models anytime — Von retains context across selections.

Von’s routing is optimized for simplicity and efficiency. Use manual selection when you have a clear reason to override the default.


The Model Selector offers flexibility. Von’s smart routing provides simplicity. Use them together to work faster, smarter, and more cost-effectively.

Was this helpful?
Copyright © Von.dev. All rights reserved.