Skip to content
Moka

Compare models

Three models answer the same prompt with the same MCP tools, side by side

Compare (top bar) sends every prompt to 2–4 lanes at once. Each lane uses a different model but the same workspace: same MCP servers, skills and system prompt — so the comparison is fair.

Per lane you get:

  • ⏱ total time and time-to-first-token
  • 🪙 tokens used
  • 🔨 number of tool calls
  • the full conversation, including tool calls and generative UI

Lanes keep their own history, so follow-up questions work. Change a lane’s model to reset it; Clear resets all lanes. Every run also appears in the inspector.

Typical uses: choosing a model for a use case, checking a cheaper model still calls tools correctly, or demoing the difference between a local and a frontier model.