BREAKING
Moonshot AI Unveils Kimi K3
0
T
total parameters
0
experts
0
active per token
0
context tokens
0
$
per M input
0
$
per M output
Kimi K3 Coding Benchmarks
Program Bench
77.8
SWE Marathon
42
Closing the Gap, But Not Closed
Kimi K3
China
●
Leads on coding tests
●
Open weights around July 27
●
Cheaper API pricing
U.S. Frontier
Claude / GPT
●
Ahead on aggregate benchmarks
●
Claude Fable 5
●
GPT 5.6 Sol
Full Weights Will Tell the Story
AI NEWS BLITZ
Chinese startup Moonshot AI has released Kimi K3, reigniting the U.S.-China AI race.