DeepSeek 4.1 Flash 0910 is perfect for M5 Ultra 256GiB. Running it fully resident in RAM, prefill at ~2500 tok/s and decode at ~40 tok/s. Probably tons of room to improve from there.

▲

loehnsberg an hour ago | parent | next [-]

Which quantization are you using there?

	▲	liuliu 15 minutes ago \| parent [-]
		My own: https://huggingface.co/drawthingsai/DeepSeek-V4.1-Flash/tree...

▲

sdg03ksdv0d 3 hours ago | parent | prev [-]

You are running this now? :o