Skip to content
Documentation
Default Model SettingsTopic Naming Model

Topic Naming Model

Configure the model that auto-summarizes conversations into short titles. A lightweight model is recommended.

Definition: After a user starts a new chat and sends the first message, the system uses this model in the background to summarize long messages into a short title (e.g., summarize "Please draft a Friday meeting memo..." to "Friday meeting memo").

⚠️ Cost and performance warning Avoid configuring reasoning models here (e.g., deep-thinking models like DeepSeek-V4, GLM-5.2).

  • Why: Reasoning models perform long chain-of-thought before output. Using them for simple titles is overkill, making title generation very slow and unnecessarily expensive.
  • Best practices:
    • Recommended models: Qwen3.5-Flash, DeepSeek-V4-Flash, GLM-4.7-Flash.
    • Key needs: Fast, instruction-following, and cheap. It only needs to understand "summarize as a title."

Default Assistant Model

Configure the default model for new conversations, affecting initial user experience and API costs.

Translation Model

Configure the model powering selection and message translation features, affecting multilingual collaboration quality.

On this page

Topic Naming Model