About ModelYield
Task-level outcomes, explicit identity and honest cost.
An independent measurement lab
What this site is
ModelYield is an independent measurement lab for AI coding work: what it costs to get a task actually done, on which execution path, with the failures left in the denominator. It is maintained by a single maintainer and is not affiliated with any model provider, router project, or benchmark listed in the sources.
Current state, stated plainly
No paid benchmark has been published yet. The pages currently ship a clearly-labelled synthetic example suite and a cost calculator for your own numbers. When real measurements exist they will carry their own identity, dates, and limitations — the same honesty rules this page is written under.
Editorial independence
Any future sponsorship or affiliate relationship is disclosed on-page and cannot influence candidate selection, task design, ordering, or corrections. Measurements cite their raw attempt ledgers.
Response expectations
Write via the contact page; replies are batched, normally within two weeks. There is no vendor support service here.