Comment on Show HN: Smart model routing directly in Claude, Codex and CursorComments−virangjhaveri2moDo you reward the RL model based on the token consumption when multiple LLMs complete the task ?−adchurchOP2moEffectively yes (based on cost though, not raw token count)
Comments
Do you reward the RL model based on the token consumption when multiple LLMs complete the task ?
Effectively yes (based on cost though, not raw token count)