Replying to @LMenardBlondin
And then there's whatever the hell Deepseek is doing. We're currently on deepseek-v4.1-flash(-1017), they're really out here releasing substantial updates behind the same model name and they barely even let consumers know. You gotta be tracking HF or the API. 💪💪🔥
Replying to @tech234a
That was my first approach I had working the other day, and that's ported! Besides minor performance improvements, also been needing an excuse to dabble with WebGPU






