Anyone (developers, procurement teams, the neurodivergent community) can
see how systems compare. Vendors can respond on the record. Below is a{" "}
preview with mock data; the live
leaderboard launches with the first cohort.
{/* preview banner */}
PREVIEW · mock data · live launch with first cohort, Q3 2026
);
}
function MethodologySection({ theme }) {
const { Section, Eyebrow, Placeholder } = window;
const { isMobile } = window.useViewport();
const [open, setOpen] = React.useState(0);
const items = [
{
q: "How is independence kept?",
a: (
<>
Philanthropic funding sits in a separate operating account from Benchmark
scoring. The team that runs evaluations does not see funder identity for
systems under test.
),
},
{
q: "Who sets the criteria?",
a: (
<>
The BRAINS Trust, a standing committee of neurodivergent clinicians,
researchers, ethicists, educators, and policy professionals, owns the
rubric. Criteria changes go through public consultation and a Trust vote.
),
},
{
q: "How are evaluations reproduced?",
a: (
<>
Every evaluation publishes the pinned system version, the exact prompts,
the quantitative test harness, and a representative sample of Trust
transcripts. An external lab can run the same harness against the same
version and arrive at the same score, within a published tolerance.
),
},
{
q: "How do vendors appeal a score?",
a: (
<>
Vendors can request a re-evaluation when they ship a material change to the
system under test. Disputes go to a Trust sub-panel that did not run the
original evaluation; the appeal record is published whether or not the
score moves.
),
},
{
q: "Does BRAINS evaluate its own products on the same terms?",
a: (
<>
Yes. BRAINS Incubator products and BRAINS partner products are evaluated on
exactly the same rubric, by the same Trust panel rotation, and shown on
the same leaderboard. Conflicts of interest are disclosed inline on the
system's profile.
),
},
];
return (
Methodology & independence
The score only holds up if the{" "}
method is open.
Benchmark is a product of the BRAINS Incubator, and it evaluates the
Incubator's other products, and our partners' products, on exactly the same
terms as any other AI system.