This post has been de-listed
It is no longer included in search results and normal feeds (front page, hot posts, subreddit posts, etc). It remains visible only via the author's post history.
561
Google is about to release an o1-style reasoning model - "centaur" on the LMSYS Arena gets one of my hardest benchmark questions consistently correct, *without showing any work or "thinking" in its output*, but takes roughly 30 seconds to stream the first token
Post Flair (click to view more posts with a particular flair)
Author
Account Strength
100%
Account Age
5 years
Verified Email
Yes
Verified Flair
No
Total Karma
40,536
Link Karma
34,355
Comment Karma
6,100
Profile updated: 9 hours ago
Subreddit
Post Details
We try to extract some basic information from the post title. This is not
always successful or accurate, please use your best judgement and compare
these values to the post title and body for confirmation.
- Posted
- 1 month ago
- Reddit URL
- View post on reddit.com
- External URL
- i.redd.it/h9mwnljmun6e1....
Something else really cool. Take a model that can write an entire application, start to finish with testing. Double compute and memory resources and it does it in half the time. Keep doubling and eventually it's writing an entire application faster than you can prompt it.
Of course as the hardware gets better so does the software. Gemini 2.0 flash is faster and better than many slower models. Speed increases happens from both sides right now.