bellwethr CENSUS · DATA SUBMISSION

The inference census

Help build a single, standard price index for AI inference — in the open. Tell us what you're paying, how much you use, and how you think the index should be calculated. Everything is optional and aggregated into open research data.

01 Ideas, critiques & contact
DATA OR THINKING WE SHOULD KNOW
EMAIL — TO JOIN THE WORKING GROUP (optional)

Prefer to email us directly? Reach the census team at survey@bellwethr.org

02 About you
WHICH BEST DESCRIBES YOU?
03 Your usage
APPROXIMATE MONTHLY TOKEN VOLUME
WHICH PROVIDERS DO YOU USE? (select all)
04 Methodology

The core debate: how should BWI actually be calculated? Pick the option closest to your view on each — or skip.

CAPABILITY GATE — which benchmarks qualify a model?
THRESHOLD — how good must a model be?
BASKET — how to combine prices into one number?
REFRESH — how often should the index update?
05 Prices you pay

Add the per-million-token prices you actually pay (negotiated or list). This is the raw signal the index is built from.

PROVIDER MODEL $ / 1M IN $ / 1M OUT
all fields optional