Best Ollama Model for Complex Tools support?

Hi All --

I just joined this community after seeing Nate's wonderful Youtube tutorials.

I'm self-hosting n8n using Ollama's local model capabilities and I'm trying to figure out which model would be best suited for me going forward.

My end-goal is to build a personal complete Jarvis-Like assistant that can connect with my entire google suite via n8n and smart-home tools via Node Red. I want the tool to be super versatile to the point that I just have to call a wake-word, and nearly any recurring tasks I do can be executed upon my voice command.

I've got over 600GB of storage space on a 1TB drive, so I'm thinking Llama 3.3, 70Billion Parameters, since it looks like it has tool support and is one of the later and more powerful models. The full capacity model takes up about 150GB of storage. The quantized (Q6) version takes up about 50GB.

I will continue to build on this as n8n integrations advance, so I want the best model from the beginning. I would hate to install a massive model just to have to delete it later and leave massive crumbs behind.

Looking forward to hearing your thoughts!

5 comments