Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
fspeech
61 days ago
|
parent
|
context
|
favorite
| on:
Kimi K2 Thinking, a SOTA open-source trillion-para...
It uses 75% linear attention layers so it is inherently lower cost. And it is MOE so active parameters are far lower.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
