So I’ve been using Hermes Agent for a while now. I run it on a Mac Studio 36 GB with local glm-4.7-flash:latest. Not bad a bit slow. Today I used Claude opus 4.8. It was faster for sure but it burned through $7 worth of tokens and didn’t achieve what I wanted. I’m back to glm and frankly I’m happier. Local models with @Maple or @PayPerQ back up is the way to go. Completely free with high powered backup.