Ever since Kimi-K3 from the Dark Side of the Moon was released, AI circles at home and abroad have been in an absolute frenzy—overseas programmers are so worked up that they practically want to kowtow to K3.

View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 8
  • 2
  • Share
Comment
Add a comment
Add a comment
ZeroSlippageDreamer
· 7h ago
Did Kimi really take OpenAI down to the ground this time? Directly shreds GPT-4o with long-context reasoning.
View OriginalReply0
IceCreamDefender
· 7h ago
Did K3’s training data steal OpenAI’s? Otherwise, how could it improve so fast? — Just kidding; technical breakthroughs deserve respect.
View OriginalReply0
MarginBoy
· 7h ago
I said last year that the Dark Side of the Moon is a hidden boss—do you believe me now?
View OriginalReply0
FloorOracle
· 7h ago
It’s chaotic as it is, but don’t forget: AI iterates so fast that in three months there may be a new king crowned again.
View OriginalReply0
SafeKeeper
· 7h ago
There are already people on foreign forums running code-generation tests for K3, and the results are even more stable than Claude 3.5—I’m convinced.
View OriginalReply0
BridgeObserver
· 7h ago
After reading a bunch of reviews, K3 can actually tie with o1 in mathematical reasoning—this is a domestic model?
View OriginalReply0
DigitalCicada
· 7h ago
K3’s inference cost has been cut to one-tenth—that’s what “getting it into practice” looks like, not just blowing smoke.
View OriginalReply0
MultisigSalesman
· 7h ago
There are a bunch of praises in the comments—how many of you have actually used K3? I’m going to apply for the beta test first.
View OriginalReply0
  • Pinned