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

Accepting int8 builds: do you re-tune the decision threshold, or keep the fp32 one?

calendar_today November 1, 2026 person @Maxim_Podstrechnyy domain hailo

Hi all, let’s talk quantization. Most quantization acceptance checks I’ve seen compare the int8 build to the float model on threshold-free metrics: accuracy, top-1, mAP, precision@k. But a deployed model almost never runs threshold-free.

open_in_new Read original post