so far yah. fast too. I havent done any fine tuning yet but it looks like will make a great base model for agents. It may replace stewart/multi-agent:7b in my toolkit