I'm using this model right now in claude code with LM Studio perfectly, on a macbook pro

You mean Qwen3-Coder-Next? I haven't tried that model itself, yet, because I assume it's too big for me. I have a modest 16GB MacBook Air so I'm restricted to really small stuff. I'm thinking about buying a machine with a GPU to run some of these.

Anywayz, maybe I should try some other models. The ones that haven't worked for tool calling, for me are:

Llama3.1

Llama3.2

Qwen2.5-coder

Qwen3-coder

All these in 7b, 8b, or sometimes 30b (painfully) models.

I should also note that I'm typically using Ollama. Maybe LM Studio or llama.cpp somehow improve on this?

I’m mostly out of the local model game, but I can say confidently that Llama will be a waste of time for agentic workflows - it was trained before agentic fine tuning was a thing, as far as I know. It’s going to be tough for tool calling, probably regardless of format you send the request in. Also 8b models are tiny. You could significantly upgrade your inference quality and keep your privacy with say a machine at lambda labs, or some cheaper provider, though. Probably for $1/hr - where an hour is a many times more inference than an hour on your MBA.

I've not used Ollama in a long time, but I believe it aggressively quantizes models by default, leading to subpar performance.

Surely the answer is a very small proxy server between the two?

That might work, but I keep seeing people talk about this, so there must be a simple solution that I'm over-looking. My solution is to write my own minimal and experimental CLI that talks JSON tools.