Kimi K3 Beats GPT-5.5 in First-Pass Game Draft

ChrisGPT · x · 2026-07-18

The author tested Kimi K3 Standard on the initial draft of their indie space game, noting its first output was noticeably better than GPT-5.5's first draft, calling it an open-source model at the "GPT-5.5 to GPT-5.6 level."

They also shared a cost breakdown: this run cost an estimated $2.50 in tokens, compared to $22 for Fable 5 and $14 for GPT-5.5. However, they added that Fable 5 produced more "visually stunning" graphics, though the overall gap wasn't that wide.

Related event: Kimi K3 Beats GPT-5.5 in Game Generation Test(2 posts)→

Original post →

More from Models

Models channel →