Skip to content

Akros · Accuracy

We measure ourselves in public.

Akros has the benchmark harness for measured food-photo accuracy, but no public Akros MAE is published until source-tracked image rows, ground-truth weights, and Akros estimates are attached. We do not run head-to-head comparisons against competitors that do not publish per-meal data; we cite their published ranges instead.

This week’s Akros runs

No Akros food-photo accuracy number is published yet.

The launch gate requires at least 25 source-tracked meal images with measured ground truth and Akros estimates before this table can show an Akros MAE. Synthetic contract rows and USDA/FNDDS ground-truth rows validate the harness, but they are not public photo-accuracy evidence.

Published competitor ranges

We do not fabricate Cal AI vs Akros per-meal comparisons because Cal AI does not expose per-meal accuracy data. Below are the figures each competitor (or independent reviewer) has published — sourced and linked.

How we run the benchmark

The benchmark runner accepts source-tracked meal images, measured kcal/macros, and the same food-photo endpoint your iPhone calls. It records returned macros, latency, p50/p95 kcal error, macro error, and row-level failures against the labelled truth.

Only measured rows with image provenance and attached Akros estimates can be published here. We refuse to scrub failed rows from a published run. If a release regresses accuracy, the next row should make it visible before we tell ourselves it is fine.

Full methodology — dataset licence, exclusion rules, confidence intervals, vendor comparison limits — lives at akros.life/accuracy/methodology.

Akros is a personal wellness app. It is not a medical device, does not provide medical advice, and is not a substitute for consultation with a licensed clinician.