Blog
Kimi K3 vs Fable vs Sol on a Normal Debug: The Cheapest Model Wasn't Cheap
A Tailscale login kept firing on my Mac every 15 minutes and I thought I was compromised. So I turned Fable 5, GPT-5.6 Sol and the cheap Kimi K3 loose on the same triage. Kimi K3 looks cheap per token, but driven by outcomes it lands in the same price band as the pricey two, and takes twice as long. The lesson: price your models by the outcome, not the token.