See the models used at each review depth.
Light, Standard, and Premium do more than change response length. Each depth uses a different model lineup and a different output and validation budget.
Active server configuration
These names come from the model configuration currently used by the API server, not a hard-coded marketing list. The lineup may change after a verified rollout or rollback. Each completed result also records the models used for that specific run.
Active model lineup
Loading the active model lineup.
What changes with each setting?
- Review depth
- Light, Standard, and Premium change the primary model lineup and the output and validation budget.
- Rounds
- 2R focuses on the core exchange. 3R adds a closing rebuttal and final critique for another pressure test.
- Participating AIs
- 2A uses GPT and Claude. 3A adds Gemini for a third-angle review and final check.
DDT is not the price of one model call.
DDT covers staged cross-model calls, debate-state summaries, quality validation, retry headroom when needed, result assembly, and retention. Total review work also changes with the number of rounds and participating AIs.
A higher review depth does not guarantee a correct answer. For important decisions, review the hidden assumptions, undefended claims, and evidence that could change the judgment.