BREAKING
Microsoft Tests Kimi K3 for Copilot
0
T
parameters
0
M
token context
0
of 896
experts per token
API pricing per million tokens
Input
3
Output
15
Cached in
0.3
Coding wins vs lingering gaps
Strengths
1,679 pts
●
Led LMArena Frontend Code Arena
●
Beat Claude Opus 4.8, GPT-5.5 on some
●
Praised for front-end and visual work
Limits
●
Trails Claude Fable 5, GPT-5.6 Sol
●
GPU capacity constraints
●
429 rate-limit errors reported
Kimi's path into Microsoft
1
Kimi K2.6 in Foundry
↓
2
K2.7 Code in GitHub Copilot
↓
3
K3 under evaluation
Vendor economics shifting
AI NEWS BLITZ
Microsoft is said to be evaluating Moonshot AI's Kimi K3 model inside its Copilot products.