FlappyBench Showdown: DeepSeek V4.1-Flash Wins on Cost, GLM 5.3 on UI, Kimi K3 Lags

Command Code ·

Key Info

A FlappyBench run using the same /design prompt compared three models: DeepSeek V4.1-Flash, GLM 5.3, and Kimi K3, with notable differences in both quality and cost.

Highlights

  • DeepSeek V4.1-Flash wins on cost-efficiency: scored 9/10 with a one-shot cost of $0.0089, the lowest of the three.
  • GLM 5.3 delivers the best result: scored 9.5/10 and nailed the UI, but cost about 2x more than DeepSeek at $0.0184.
  • Kimi K3 trails the field: scored 8/10, used hard-coded gameplay, and cost roughly 8x more at $0.0740.
Loading...