kimi-k3 vs gpt-5.6-luna
Pricing, Performance & Features Comparison
Moonshot’s 2.8T-parameter flagship for long-horizon coding and knowledge work, with native vision and a 1M-token context window; thinking is always on.
Input$3
Output$15
Latency (p50)5.8s
Output Limit1M
Function Calling
JSON Mode
InputText, Image, Video
OutputText
in$3out$15cache$0.3–
Latency (24h)
Success Rate (24h)
Small tier of the GPT-5.6 family (Luna), the fastest and most cost-efficient GPT-5.6 model.
Input$1
Output$6
Latency (p50)4.1s
Output Limit128K
Function Calling
JSON Mode
-
InputText, Image
OutputText
in$1out$6cache$0.1–