BREAKING
Kimi K3 Tops GPT-5.6 & Fable 5
Game-Design Test Scores
Kimi K3
9.5
Fable 5
7.5
GPT-5.6 Sol
7
0
$
Kimi K3 per run
0
$
Fable 5 per run
0
$
GPT-5.6 Sol per run
0
x
cheaper than Fable 5
0
x
cheaper than GPT-5.6 Sol
Why It Matters
Kimi K3
Moonshot
●
2.8T parameters, 1M context
●
Weights set to go open
●
Strong design and front-end
Caveats
context
●
Inference can lag
●
Narrow single-domain test
●
Rivals lead on agentic work
Premium Pricing Under Pressure
AI NEWS BLITZ
Moonshot's Kimi K3 beat two American flagships in a head-to-head game-design test.