[jki275]

I can definitely write code with a local model like Devstral small or a quantized granite, or a quantized deep-seek on an M1 Max w/ 64gb of ram.