Skip to content
Browse all documentation

Build · 6 min read

Models and reasoning effort

Choose a Gyroscape model and effort level that fits the task, latency, and usage.

On this page
  1. Use the model picker
  2. Choose an effort level
  3. What Ultra means
  4. Change settings with the task

Use the model picker

Open the model control to see what the account can currently use. GyroThink and GyroThink Pro are Gyroscape's primary product models. Additional served models can appear by plan, deployment, and availability.

Changing models does not change the directory or Gyroscape tool and permission system. It can change reasoning behavior, latency, context support, and usage.

Choose an effort level

Default fits most work. Lower effort can suit small obvious changes. Higher effort helps with unfamiliar repositories, multi-file changes, debugging, migrations, and verification-heavy work.

Only levels supported by the selected model appear, and the setting is preserved per task where supported.

What Ultra means

Ultra is a Gyroscape orchestration mode on supported GyroThink models. It combines the strongest supported reasoning with proactive, return-on-investment-driven delegation and deeper verification.

Ultra does not force multiple agents. It can use zero parallel lanes when delegation would duplicate work. Expect higher usage and possibly longer execution.

Change settings with the task

Use a faster setting for a narrow follow-up and a stronger one for diagnosis or architecture. If a model is temporarily unavailable, keep the task and choose another available model.