I’d like to see this extended into a full automatic model-routing mode in ChatGPT and Work.
Rather than requiring the user to manually choose Luna, Terra, Sol or Astra, ChatGPT should be able to select the cheapest model that can reliably complete each stage of a task, then automatically escalate when complexity, uncertainty, confidence or risk requires a stronger model.
For example:
- Luna for repetitive/high-volume work
- Terra for normal implementation and analysis
- Sol for complex reasoning and architecture
- Astra for the hardest audits, debugging and multi-step workflows
The model should be able to change during the same conversation or task while preserving full context.
User-selectable modes could include Economy, Balanced, Quality, Maximum and Manual.
ChatGPT should also display which model/reasoning level it selected and allow the user to override it.
This would be especially valuable for long-running projects where different stages require very different levels of intelligence and users currently have to keep manually changing the model selector.