BREAKING
Moonshot AI Launches Kimi K3
0
T
parameters
0
active experts
0
total experts
0
M
token context
0
x
scaling vs K2
Cheaper API pricing
$/M input
3
$/M output
15
Benchmark standing
Kimi K3
open weights
●
Tops Frontend Code Arena
●
Strong on agentic coding
●
~2.3x lower cost
US rivals
closed
●
Fable 5 leads some indices
●
GPT-5.6 Sol ahead overall
●
Faster in some cases
Full weights due July 27
AI NEWS BLITZ
China's Moonshot AI has unveiled Kimi K3, calling it the largest open-weight model ever built.