← sources

Possible to run Qwen3.8-Flash-Next-4bit on Mac with 64GB?

source: reddit (reddit:LocalLLaMA)published 2026-10-10 18:33ingested 2026-10-11 02:41

retained text

Update: The answer is YES!

https://www.reddit.com/r/LocalLLaMA/comments/1x2tstj/omg_if_you_have_a_mac_with_64gb_try/

I have m3Max 64GB, and I can allocate up to 58GB to GPU.

Is that possible to run Qwen3.8-Flash-Next-4bit? I'd like to avoid lower quant than 4bit.

I know 58GB can't load everything, but I think there are now some engines that let you load only some parts of the model and stream from SSD?

If so with what engine: Strata, omlx?

Also roughly how much context can I expect to run, and at what speed?

Thanks!

document fetched 2026-10-11 00:38 · canonical https://reddit.com/r/LocalLLaMA/comments/1x2mhso/possible_to_run_qwen38flashnext4bit_on_mac_with