Just tried jan3.ai running Llama 3B or Mistral 7B locally and it works on older intel macs that have at least 8GB of ram 

she runs hawt but its not slow by any means