Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
ein0p
6 months ago
|
parent
|
context
|
favorite
| on:
Mistral AI Launches New 8x22B MOE Model
Nope. Just the weights would take 88GB at 4 bit. 128GB MBP ought to be able to run it. If I were to guess, a version for Apple MLX should be available within a few days, for those of us fortunate enough to own such a thing.
Art9681
6 months ago
[–]
It’s already available. I had it running yesterday morning in an M3 MAX 128GB. I get about 6tps.
https://www.reddit.com/r/LocalLLaMA/s/MSsrqWHYga
Consider applying for YC's W25 batch! Applications are open till Nov 12.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
