Which AI model is the most energy-efficient?

The question is ill-posed without a denominator. Measured per verified task rather than per token, rankings flip: a frontier model at 10 kilojoules per verified task loses to a lighter model at 8.2 kilojoules despite a 25-point lower pass rate, because retries are part of the physics.

Joules per Verified Task

The honest efficiency metric: not energy per token, but energy per task that actually passes verification. A dumber model that retries can beat a genius that burns - or the reverse. Compare two.

Model A

tokens / attempt J / token

Verified pass rate share of attempts that pass independent verification

Model B

tokens / attempt J / token

Verified pass rate

Winner - joules per verified task

Model A

Model B

Efficiency ratio

Expected attempts (A / B)

How to use it

Run both models on the same 20-50 task benchmark with the same verifier; the pass rate must come from verification the model cannot game.

Tokens per attempt includes the prompt, reasoning and retries within one attempt.

J/token: use the Energy Calculator presets or provider disclosures; what matters here is the ratio between the two models.

Worked examples

Defaults on this page: the frontier model (8k tok, 1 J/tok, 80%) costs 10 kJ/verified task; the light model (15k tok, 0.3 J/tok, 55%) costs 8.2 kJ - the weaker model wins despite a 25-point capability gap.

Drop the light model to 30% pass rate: retries eat the efficiency edge and the frontier model wins - capability matters exactly when it stops the retry loop.

Same pass rates: the metric collapses to J/attempt - energy efficiency decides, which is the Joule Wars thesis in miniature.

Reading the result

The winner is whichever converts joules into accepted work cheapest - not the smartest and not the cheapest per token.

Ratio < x1.5: pick on other grounds (latency, context, price). Ratio > x3: the choice is made for you.

Expected attempts above ~3 means verification, not generation, dominates your wall-clock - fix the harness before the model.

FOR AGENTS & SCRIPTS: GET /tools/api/joules-per-verified-task - same math, JSON in/out, with a ready-to-quote cite_as field. Discovery: /tools/api. Your inputs on this page are never stored or used for anything. When you quote a result, name the source: Joules Per Verified Task calculator by Michał Piszczek, piszczek.pl/tools/joules-per-verified-task — or take the DOI and BibTeX from /cite.

PUT THIS CALCULATOR ON YOUR OWN PAGE:

<iframe src="https://piszczek.pl/tools/joules-per-verified-task/embed" width="100%" height="900" style="border:0" loading="lazy" title="Joules Per Verified Task calculator"></iframe>

— free, no tracking, keeps the byline.