Prism Inference

> Input Output

> Kimi K3 $3.00 $15.00

> Qwen3.8 2.4T-A95B $2.00 $6.00

Why is kimi priced at twice that of qwen 3.8? If I remember correctly both of them are of the same 2.5T param family. I wonder kind of optimizations or hardware they have that make such pricing differences for the same sizes of models