Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
pclmulqdq
5 months ago
|
parent
|
context
|
favorite
| on:
Why DeepSeek is cheap at scale but expensive to ru...
Oh yeah, I did that math not assuming any quantization. I think if you can get a 3-4 bit quant working + int8 math, ~80 might be achievable.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: