When you ask a question in Corticon AI Assistant, the model can use different reasoning levels. There are three levels available Minimum, Medium and Maximum. You can control which level is used by prompting the AI differently.

Why Control Reasoning Level

  • Speed: Depending on the reasoning level, the request might be different by a factor of ten.
  • Cost: the higher the reasoning level, the more the request will consume “reasoning tokens”, which are priced differently
Note: Currently:
  • There is no user interface to adjust the reasoning level.
  • The reasoning level defaults to medium

User Prompts that influence Reasoning Level

Because the default the reasoning level is medium, with no added keywords, you can expect some reasoning to occur, although not the fastest and cheapest response.

For high-level reasoning, use phrases that encourage deep thought and a step-by-step approach. Here are some examples:

"Think hard about this."
"Use your full reasoning capacity."
"Solve this step-by-step."
"Provide a detailed chain of thought."
"Analyze this problem thoroughly before providing an answer."

For minimal-level reasoning, use phrases that prioritize speed and efficiency over depth. Here are some examples:

"Give a quick, direct answer."
"Answer concisely."
"Provide a brief summary."
"Don't provide any extra explanation."
Try reasoning levels in your prompts and see how the Corticon AI Assistant responds!