Skip to content

Core References

AI Model Discovery Protocol

A four-phase protocol for sampling AI models against your own research, building task-model matches, exploring temperatures, and developing a personal strategy

The AI Model Reference Guide tells you what's out there. This page is the protocol for finding what fits you — a four-phase experiment you run against your own research.

Tip

Read this if: You have access to several frontier models (directly or via OpenRouter) and want a structured way to discover which combinations serve your research, not someone else's benchmark list.


Getting Started

Phase 1: Capability Discovery

Sample Everything (1-2 days of exploration)

  1. Access via OpenRouter: All models available through single API key
  2. Choose one complex research task (e.g., theory synthesis from 3 papers)
  3. Run the same prompt across ALL models:
    • GPT-5.4, Gemini 3.1 Pro, Claude Opus 4.6, Claude Sonnet 4.6
    • DeepSeek V3.2, Kimi K2.5, Qwen 3.5, GLM-4.5
  4. Note differences: Style, depth, accuracy, approach
  5. Test temperature variations: Try each model at 0.1, 0.6-0.8, 1.0
  6. Experiment with reasoning modes: Built-in vs. MCP Sequential Thinking

Once you've found the models and temperatures that suit your work, the natural next step is to wire them into your workspace — see the Cherry Studio Setup Guide.