What a unit of completed work costs. The denomination every buyer actually wants — and the one the market has not defined.
The Grid publishes no price per task, and will not until the market defines the unit. The per-task cost of compute is the number every buyer actually wants — which is exactly why publishing a defensible-looking wrong one would be the most damaging thing this surface could do. The rung is declined for three stated reasons.
There is no governed, market-accepted definition of “a task” to price. A price for an undefined deliverable is not a price.
Where task suites exist, they are optimisation targets; a $/task print built on them would inherit their drift without inheriting any governance.
Tokens-per-task varies on the order of 2× between models on the same work; any single $/task number silently embeds a model choice and presents it as a property of the market.
A standardised, independently governed task definition — one with the governance shape of an audited benchmark: submission, peer validation, publication under rules — would make the unit priceable. If the market produces that unit, The Grid reopens the rung by methodology change notice, with a full specification, before any figure prints. Until then, the declination stands as published methodology, dated 2026-08-20.