How AI is applied across API Evangelist and APIs.io. Read my AI disclosure →
API Evangelist API Evangelist
Discovery
Learnings
Guidance
Toolbox
Alignment
API Evangelist LLC

Tokens per second won't tell you what an answer costs on a Strix Halo

calendar_today November 2, 2026 domain ciq

On a Strix Halo, a model 13x worse per token can be cheaper per answer. Per-answer cost, speed and active parameters across 27 local models

open_in_new Read original post