Today we are shipping GeoAxis Version 3.0 — the biggest update to the GeoAxis AI location finder since we launched. V3 raises the bar for AI image location accuracy, ships a redesigned evidence engine, and — as the benchmark below shows — confirms that GeoAxis is currently the best AI location finderavailable, beating every general-purpose AI on image geolocation.
New AI image location backbone, evidence pins on every candidate, batch image geolocation up to four photos at once, and a benchmark win on Dataset-GSS over GPT 5.4, Grok 4, Gemini 3.1 Pro, and Claude Opus 4.7.
Why We Built GeoAxis 3.0
Every AI vendor wants to claim they have the best AI location finder. Most do not. We built GeoAxis 3.0 because the rest of the market kept regressing on the one thing that actually matters: finding the location of an image accurately and explainably. The old guard either guessed at country level, hallucinated a confident-sounding wrong city, or refused to answer at all.
GeoAxis V3 was rebuilt around three principles:
- Accuracy is the product. If the AI image location is wrong, nothing else matters. V3 was trained and evaluated on a stricter image geolocation rubric than we have ever used.
- Every result must be explainable. Black-box AI location finders are dead on arrival in OSINT, journalism, and insurance. V3 surfaces the visual evidence the AI used to find the location of an image.
- It must be fast. Image geolocation in seconds, not minutes. V3 is the fastest GeoAxis pipeline we have ever shipped.
The Benchmark — GeoAxis vs Every Other AI Location Finder
We ran GeoAxis V3 head-to-head against the strongest frontier AI image location models on Dataset-GSS — 615 high-quality street-level images drawn from 396 cities across 123 countries. Each model was scored on the standard 0–10 image-geolocation rubric, where 0 means “no location” and 10 means the AI image location landed on the correct spot.
The Best AI Location Finder — Dataset-GSS Image Geolocation Benchmark
Each model was scored on every image in Dataset-GSS on the standard 0–10 image-geolocation rubric — from “No location” (0) to “Correct spot” (10). Greener bars are better. GeoAxis V3 finishes ahead of every frontier model, which is why we say it is currently the best AI location finder.
Images scored (cumulative)
Methodology. We evaluated every model on Dataset-GSS, which comprises 615 high-quality street-level images from 396 cities across 123 countries. Each image was scored on the standard 0–10 image-geolocation rubric, from “no location” (0) to “correct spot” (10). GeoAxis V3 results are from our internal eval on the full Dataset-GSS set; frontier-model rows reflect their own image-geolocation performance on the same dataset.
The result is the clearest evidence to date that GeoAxis is currently the best AI location finder. V3 lands a “correct spot” result on the majority of the 615 Dataset-GSS images — more than Claude Opus 4.7, more than GPT 5.4, more than Gemini 3.1 Pro, and more than Grok 4. No frontier AI image location system comes close on this benchmark.
“When the goal is finding the location of an image, GeoAxis 3.0 has the highest rate of correct-spot answers we have measured from any AI location finder.”
What Changed Inside the AI Image Location Engine
New Vision Backbone
The V3 backbone is a larger, geo-specialised vision model trained on far more image geolocation supervision than V2. It improves AI image location accuracy on every region we tested — particularly in the Americas, Southeast Asia, and Sub-Saharan Africa, which were previously the weakest spots for every AI location finder.
Tool-Aware Evidence
GeoAxis V3 can now call external geospatial tools mid-reasoning — satellite imagery lookups, street-view alignment, signage OCR — and fold the results back into the AI image location decision. Every tool call surfaces as an evidence pin on the result map.
Batch Image Geolocation
On global search you can now upload up to four images at once. V3 performs joint AI image location across the batch, which dramatically improves accuracy when you have multiple frames of the same scene.
Indoor Geolocation via HyperVision™
Indoor Geolocation is rolling out as a HyperVision™enterprise feature. Indoor mode matches your query photo against a per-venue reference image index — the same City Search pipeline retargeted at interior spaces — and is useful for hotels, train stations, retail venues, and office campuses that ship GeoAxis their own reference imagery. General availability is still marked Coming Soon; HyperVision customers can request early access via a custom deployment.
Performance
AI image location latency in V3 scales with the Effort slider on Global Search. Low finishes in about 35 seconds with a single Gemini 3.1 Pro pass, Medium in about 50 seconds with two parallel passes, and High in about 80 seconds with four parallel passes. Across all three effort levels, V3 is materially faster than V2 at the same accuracy bucket.
Pricing and Availability
GeoAxis V3 is live today for every paying plan. Pro and Business customers get the new AI image location backbone, the Effort slider, and batch image geolocation (up to four images per Global Search run). The free public demo at geoaxis.ai/demo also runs on V3 — no account needed — so anyone can try the best AI location finder before they buy. See the pricing page for the full plan comparison.
If you are an enterprise team looking to use the best AI location finder at scale, HyperVision™ offers custom city deployments, Indoor Geolocation early access, and a private AI image location pipeline.
How to Try GeoAxis 3.0
The easiest way to experience V3 is to run the interactive demo. Upload any photo and the AI location finder will find the location of the image in seconds. If you are new to GeoAxis, start with our how-to-use guide and our guide to finding the location of an image.
What is Next
V3 is the new floor, not the ceiling. Next on the roadmap:
- General availability of Indoor Geolocation outside of HyperVision custom deployments.
- Vehicle identification — find the make, model, and approximate plate region of a vehicle in a photo.
- Video AI image location — geolocate every frame of a clip.
- Self-hosted GeoAxis for regulated industries that cannot send images to a SaaS AI location finder.
Until then — GeoAxis is currently the best AI location finder you can use to find the location of an image, and V3 is the strongest evidence of that yet. We are just getting started.
GeoAxis