Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
|
Jirach05's comments
login
Jirach05
5 days ago
|
parent
|
context
[–]
| on:
Gemini 3.1 Pro
Can anyone explain why these models decrease in performance on this "MCRC v2 (8-needle)" long context benchmark when thinking is turned on?
reply
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
reply