ChatGPT still provides the best performance, but at the cost of running in the cloud.
Mixtral works well for reasoning, but sometimes calls functions wrong. I get good performance locally on my MacBook, but it requires powerful hardware.
Gemma is very promising, it is easy to run on less powerful hardware, but its reasoning is not as good. It does seem better at calling functions than other models.
Good performance has been seen with the following models:
ChatGPT still provides the best performance, but at the cost of running in the cloud.
Mixtral works well for reasoning, but sometimes calls functions wrong. I get good performance locally on my MacBook, but it requires powerful hardware.
Gemma is very promising, it is easy to run on less powerful hardware, but its reasoning is not as good. It does seem better at calling functions than other models.