LocalBench early access

V0-alpha early access

Browser-first, explicit, and reviewable

The LocalBench MVP is designed to minimize hidden behavior. Users should understand what is accessed locally, what is measured, what may be stored, and what remains draft before they publish a run or join the waitlist. This copy is part of the early-access launch surface and should be reviewed before broader release.

Local connection only

LocalBench connects only to the runtime you choose on this machine. We do not expose your local runtime to the public internet.

No model file uploads

The app does not upload local model files. Benchmarking should only use the runtime API and the data needed to measure the run.

Benchmark data can be published by choice

Benchmark prompts, measured results, and summary metadata are the primary data submitted when you choose to publish a run.

Waitlist email is stored narrowly

If you join the alpha waitlist, we store your email address in Postgres only so we can send alpha notifications and handle removal or correction requests.

Public sharing stays opt-in

Public sharing should remain opt-in wherever practical, and a run should be reviewable before it is published.

Flags and exclusions are possible

Suspicious or low-confidence submissions can be flagged, rerun, or excluded instead of being forced into the public set.

GDPR-facing notice

Transparent choices for early-access visitors

We try to make the data flow legible before any submission happens.

If a visitor asks for waitlist removal or correction, the contact page should route that request so the address can be updated or deleted manually.

EEA and UK visitors should see the same clear choice about local access, publication, and opt-in sharing.

This page is an early-access disclosure draft and should be reviewed before broader launch.

Draft public disclosure copy for the early-access period.

Inspect the rest of the alpha disclosure set

Methodology Consent Terms Contact