- ColorCountry or published subnational aggregate
- BeaconPublished DataFast city cluster · 5+ signals
- LogoMost explored local model brand
LocalClaw Atlas · Beta
See where local AI is taking off.
Anonymous interest aggregates · 62 countries published at 5+ signals · Last 30 days · Updated 29 August 2026
Aggregate view
Select the globe before using the mouse wheel. Use plus and minus to zoom, Reset to restore the current world or regional view, and arrow keys to rotate.
LOCALCLAW
ATLAS
Latest global snapshot
The clearest picture we can publish today.
As of 29 August 2026, the United States ranks first for observed local AI interest in the LocalClaw dataset, with 1,309 of 3,337 anonymous signals recorded during the last 30 days. This is a directional view of interest, not a census of local AI users.
Worldwide regional drill-down · New
From country scale to regional detail.
Select any mapped country to reveal its administrative subdivisions. The United States drills from states into 3,144 Census counties and equivalents; China drills from provinces into 358 OCHA prefecture-level divisions; Australia drills from states and territories into 545 ABS local government areas and equivalents. A geography is colored and ranked only when DataFast directly publishes that same level at five or more signals. County, prefecture, and LGA boundaries stay neutral—visible for exploration, never assigned an estimated value.
| Rank | State | Signals | U.S. share | Relative activity |
|---|---|---|---|---|
| 01 | 516 | 39.4% | ||
| 02 | 171 | 13.1% | ||
| 03 | 63 | 4.8% | ||
| 04 | 58 | 4.4% | ||
| 05 | 58 | 4.4% | ||
| 06 | 43 | 3.3% | ||
| 07 | 43 | 3.3% | ||
| 08 | 29 | 2.2% | ||
| 09 | 27 | 2.1% | ||
| 10 | 22 | 1.7% |
Do not read this as residence: region and city are approximate network locations. Oregon is kept visible but explicitly flagged because 511 of its 516 signals resolve to The Dalles; that concentration may reflect network or data-center routing. Its beacon uses the approximate GeoNames network-city centroid, never an individual visitor location.
Country ranking
Where interest in local AI is concentrated.
The ranking below uses raw observed interest signals. Select a country to bring it into focus on the globe. Country rows below five signals are withheld before public JSON publication to avoid over-reading tiny samples. Within a published country, signals outside a five-or-more-signal DataFast city cluster remain visible only through the country color.
| Rank | Country | Signals | Share | Relative activity |
|---|---|---|---|---|
| 01 | 1,309 | 39.2% | ||
| 02 | 224 | 6.7% | ||
| 03 | 209 | 6.3% | ||
| 04 | 118 | 3.5% | ||
| 05 | 117 | 3.5% | ||
| 06 | 85 | 2.5% | ||
| 07 | 66 | 2.0% | ||
| 08 | 66 | 2.0% | ||
| 09 | 65 | 1.9% | ||
| 10 | 62 | 1.9% | ||
| 11 | 57 | 1.7% | ||
| 12 | 57 | 1.7% | ||
| 13 | 46 | 1.4% | ||
| 14 | 44 | 1.3% | ||
| 15 | 41 | 1.2% | ||
| 16 | 40 | 1.2% | ||
| 17 | 38 | 1.1% | ||
| 18 | 30 | 0.9% | ||
| 19 | 27 | 0.8% | ||
| 20 | 27 | 0.8% |
Install paths · New
Which model paths and setup destinations visitors selected.
The model path is attributed only when an eligible LocalClaw model page immediately precedes a supported setup selection in the anonymous journey. Destination rows identify the repository or app selected. Every published row independently reaches five unique visitors.
| Rank | Published path | Visitors | Type | Relative intent |
|---|
Read this precisely: these are selected model paths and setup destinations, not installed models. LocalClaw does not yet receive proof that a download completed, an installation succeeded, or an inference ran.
Methodology · Version 1.0 beta
A useful signal, with a visible boundary.
LocalClaw Atlas is being built in public. The beta starts with a transparent measure we actually have: anonymous, aggregated interest in LocalClaw by country, with worldwide administrative drill-downs and privacy-thresholded regional values where DataFast can publish them. Verified installations and active local model runs will only appear after a separate, opt-in telemetry system exists.
What counts today?
A country-level unique website visitor reported by DataFast during the 30-day window. Where DataFast supplies a country-filtered region breakdown, Atlas may also publish a subnational aggregate. Country, subnational, and city-cluster counts are separate provider breakdowns; city clusters are never used to derive region, county, prefecture, or LGA totals.
What does not count?
No installation, model launch, prompt, inference, machine identity, or claim about the total number of local AI users is included in this beta.
How are locations ranked?
Countries and published regional rows are ordered only from their own DataFast breakdown. No city coordinate, polygon area, population share, or parent remainder is used to estimate a region. DataFast currently has no county, prefecture, or local-government-area dimension, so those deeper boundaries are never colored or ranked. A beacon appears only for a separate published DataFast city cluster with at least five signals, at an approximate network-city centroid from GeoNames.
How is privacy handled?
The five-signal threshold is applied independently to country, subnational-region, and city-cluster rows before public JSON publication. Rows below five are omitted; only aggregate metadata is retained. Public outputs never contain individual people, devices, exact IP addresses, or exact visitor coordinates.
Source and coverage: aggregated anonymous LocalClaw website traffic measured by DataFast, 31 Jul–29 Aug 2026, Europe/Zurich. Global country boundaries use Natural Earth 1:50m; worldwide internal subdivisions use Natural Earth 5.1.1 at 1:10m. Deeper U.S. county boundaries use the official U.S. Census Bureau 2025 Cartographic Boundary File at 1:5m. China prefecture-level divisions use the OCHA/HDX COD-AB 2020 dataset, licensed under CC BY 3.0 IGO; that source explicitly describes its China boundaries as unofficial, with unknown establishment date, and no endorsement is implied. Australian LGA boundaries are based on the official ABS ASGS Edition 3 Local Government Areas 2025, licensed under CC BY 4.0. ABS states that LGAs are statistical Mesh Block approximations of gazetted boundaries, not official legal boundaries. Natural Earth sometimes mixes administrative levels, so boundaries are cartographic references rather than a claim of identical constitutional status. Boundary display and naming are not a statement on sovereignty. Published beacon centroids use approximate network-city locations from GeoNames, licensed under CC BY 4.0. Source coverage is biased toward people and network routes that reach LocalClaw.
Download country JSON Download regional JSON Download deeper-boundary manifestQuestions, answered plainly
What the index can—and cannot—tell you.
Which country leads the current snapshot?
The United States leads this 30-day LocalClaw interest snapshot with 1,309 signals. That result reflects this dataset, not all local AI activity worldwide.
What does “local AI” mean here?
AI models designed to run on personal devices or privately controlled hardware instead of relying entirely on a third-party inference API.
Which local AI models are people exploring in each country?
Open the Models view to see the leading local AI brand for each observed country and region, then select its logo for model-level detail. The ranking starts with the first visitor to LocalClaw model pages and does not measure verified model use.
Is this a local AI adoption ranking?
Not yet. Adoption requires trustworthy installation or active-use signals and a defensible population denominator. The beta deliberately reports interest instead.
What comes next?
With explicit opt-in telemetry, Atlas can add verified installed and active views, then compare raw activity with a separate per-capita measure.
Can I explore states, provinces, and federal subjects?
Yes. Select any mapped country directly on the globe. In the United States, select a state again to open counties; in China, select a province-level region to open prefecture-level divisions; in Australia, select a state or territory to open its local government areas where the official source has real additional features. These deeper boundaries load on demand and remain neutral because DataFast does not publish county, prefecture, or LGA totals. Beacons remain separate city clusters, never verified residence or individual locations.