Groundtruth vs Mireye

For developers comparing location-data APIs. Mireye is a location-data API for AI agents. Both products serve the same 351-field public-data catalog for US coordinates. This page compares only what we measured or what Mireye publishes. Mireye is a trademark of its owner; we use the name only to identify the product we compare against.

How we measured

We saved Mireye's free Fetch responses at 5 test locations (Willis Tower in Chicago, rural Iowa, Miami Beach, Paradise CA, and San Francisco City Hall), 10 field groups each: 1,685 field values. Then we compared our output for the same points, field by field. 1,289 values match within tolerance, and 343 are correctly empty on both sides. The rest are below.

Topic	Groundtruth	Mireye
Fetch catalog coverage	351 / 351 fields, plus 4 extra	351 fields
Values where Mireye errored e.g. HUC-12 watershed name, NHD area, groundwater depth	16 answered	16 errors (timeouts)
Values where Mireye was empty wind power density, Weibull k, prevailing direction at 5/5 points	16 answered	16 absent
Errors turned into a clean 'absent' with a reason	16	16 errors
Batch size per call	up to 100 locations, 8 in parallel	up to 25 locations, 4 in parallel
Per-field provenance flags	opt-in extras: vintage, as_of, substitute, fallback, cache_hit	source, confidence, fetched_at, vintage
EIA electricity / gas prices	2026-08 / 2026-07	2026-04
ACS 5-year	2020-2024	2019-2023
PM2.5 attainment	2012 annual and 2006 24-hour standards	2012 annual standard only
Wind shear 100-160 m	returned as a field	not in catalog
Outside NREL coverage (e.g. Alaska, Hawaii)	labelled fallback (PVGIS, NASA POWER, Global Wind Atlas)	no answer
Parcel enrich, parcel search, driving times, Ask Mireye is ahead here	not offered yet	offered (paid)
Landslide susceptibility parity Mireye serves a re-encoded grid we can't reproduce	matches at 3 of 5 test points	n/a

Details

Values Mireye couldn't return

huc_12_name and intersects_nhd_area timed out on Mireye at all 5 points. We answered all 5, e.g. Chicago "South Branch Chicago River-Chicago Sanitary and Ship Canal".
nearest_groundwater_well_depth_to_water_m timed out on Mireye at 4 points. We answered all 5.
Wind power density, Weibull k and prevailing direction at 100 m are listed by Mireye but returned empty at every test point. We compute them from NREL's WIND Toolkit 2007-2013 data, e.g. Iowa 433.1 W/m².

Correctness

At Miami Beach, Mireye returns a central-Pacific marine monument as the protected area. That's a polygon that crosses the antimeridian. We return no protected area, which is correct there.
Mireye's PM2.5 status covers only the 2012 annual standard. We also report the 2006 24-hour standard: San Francisco is "Nonattainment / Moderate" (Bay Area) and Paradise is "Maintenance / Moderate" (Chico).

Freshness

By default we serve the newest upstream releases. Teams that need to reproduce an older snapshot exactly can pin the vintage per source.

Where Mireye is ahead

Mireye sells parcel Enrich, parcel Search, driving-time proximity and a natural-language Ask endpoint. Groundtruth doesn't offer these yet.
Mireye's Fetch is free and self-serve today. Groundtruth API access is by request during early access.
Landslide susceptibility matches at only 3 of 5 points, because Mireye serves a re-encoded grid we can't reproduce.
Mireye lets customers request new fields.

Speed (our measurements)

Cached lookup, any field: ~5 ms per source
Same point after an API restart: 0.04 s
WIND Toolkit fields from the US-wide precompute: 27-32 ms
Power-grid fields from the local index: 8-47 ms
All 355 fields at a brand-new point: 30.1 s, 0 failures
Same, with a 5 s latency budget: 15.1 s

We haven't benchmarked Mireye's latency side by side, so we don't claim a speed comparison.

Building on location data? See the developer docs. Buying land? Check a location.