Replace ChatGPT With a Local Model — Complete Workshop April 26

The best local model to replace ChatGPT depends on your hardware and use cases. In this workshop the instructor walks you through selecting from models available through Docker Model Runner — including Llama, Mistral, Phi, and Gemma — and testing them for your specific tasks. For most personal assistant use cases a 7B to 8B parameter model works well on 16GB RAM.

Building your own local AI assistant does not prevent you from using ChatGPT. Many developers use both — local for everyday and sensitive tasks, cloud for complex tasks requiring maximum capability. This workshop gives you the option to replace ChatGPT for daily use while keeping the flexibility to use other tools when needed.

This live 4-hour workshop takes you through the complete replacement — local model running through Docker Model Runner, connected to OpenClaw, deployed to WhatsApp or Telegram. By the end of the session you have a working local model alternative ready to use.

OpenClaw maintains context within conversations and the instructor covers its memory configuration during the workshop. For long-term memory across sessions, the workshop covers the OpenClaw architecture so you understand how to extend this capability.

Local models are generally slower than cloud APIs, have smaller context windows than the latest frontier models, and may handle very complex reasoning tasks less well. For the majority of everyday assistant tasks — writing, summarising, coding help, Q&A — the difference is minimal. The instructor covers realistic expectations during the session.

Yes. Docker Model Runner works on Mac including Apple Silicon. Apple Silicon Macs benefit from unified memory which makes local LLM inference particularly efficient. The instructor covers Mac-specific setup considerations during the live workshop.