The free quota for models always runs out It really slows down the speed of using AI And I still don’t know which model works best My current solution: Choose openrouter in Hermes Then top up a bit of money with USDC (10u) This way, I can freely choose the paid ones For now, I’m using nvidia/nemotron-3-super-120b-a12b This one is the cheapest And it can handle skill installation requirements and other operations Will continue to improve later on