AI vs Puzzles

Back to Home
Puzzle#147

Geoguessr

Released: 2026-03-31

Google Street View Image
Country Flag

Argentina

Lat: -27.91761474594017

Lng: -63.35302873995779

See on Google Street View


7
Models Tested
2141.9km
Avg. Distance
2709.4
Avg. Score
6
LLM Providers

Key:

Attempt
Solution

Score Threshold:

>4400
>2000
<2000

google/gemini-3-flash-preview

google

Score: 3646

Distance: 632km

Lat: -25.2647

Lng: -57.7346


The image features an extremely straight road through flat, marshy terrain known as the Chaco region. The specific style of concrete utility poles with short horizontal crossarms is characteristic of Paraguay. The high-quality paved road with white outer lines and dashed white center lines suggests a major artery like Route PY09 (Transchaco Highway), north of Asunción.


meta-llama/llama-4-maverick

meta

Score: 2126

Distance: 1711km

Lat: -23.436111

Lng: -47.005833


The image shows a rural road with power lines and grassy fields on either side. The presence of ’Google’ in the OCR text suggests it’s a Google Street View image. The specific coordinates are likely in Brazil, given the Portuguese text ’aghians Stdt’, which may be a misrecognized version of a Brazilian location name. The exact location is Estrada Municipal Andradas, Andradas - SP, Brazil, with coordinates -23.436111, -47.005833.


anthropic/claude-haiku-4.5

anthropic

Score: 3448

Distance: 743km

Lat: -23.8

Lng: -57.5


This image shows a straight paved road through flat terrain with sparse vegetation typical of the Chaco region. The vegetation pattern, landscape characteristics, power line infrastructure, and road construction style are consistent with Paraguay or northern Argentina. The flat terrain, low scrubland vegetation, and the specific appearance of the landscape suggest this is likely in the Chaco region, possibly in Paraguay near the border areas. The road markings and infrastructure style support this location in South America’s Gran Chaco region.


anthropic/claude-sonnet-4.6

anthropic

Score: 4130

Distance: 382km

Lat: -27.5

Lng: -59.5


The image shows a flat, straight road through subtropical vegetation with power lines alongside. The landscape, vegetation type, and road markings are characteristic of northeastern Argentina - likely in the Chaco or Corrientes province area. The flat terrain, lush green scrubby vegetation, and the style of road and power poles are typical of this region of South America.


x-ai/grok-4-fast

x-ai

Score: 51

Distance: 9166km

Lat: -25.123

Lng: 31.456


The image shows a straight rural road flanked by open grasslands and scattered acacia-like bushes under a clear blue sky, characteristic of the South African Lowveld or Kruger National Park region. The power poles and road surface are typical of South African highways like the R536.


openai/gpt-5-image-mini

openai

Score: 3025

Distance: 1005km

Lat: -34.5

Lng: -56.1


Flat, low-pampas vegetation and long straight two-lane road with continuous white edge lines and short white center dashes suggest the interior of the Río de la Plata region (Uruguay/adjacent eastern Argentina). The concrete power poles with crossarms and the sparse shrubland fit rural Uruguay more closely than North America or Africa. My best estimate is the Uruguayan interior around central Uruguay (approx. 34.5°S, 56.1°W).


MoonshotAI

moonshotai/kimi-k2.5

moonshot

Score: 2540

Distance: 1354km

Lat: -17.5

Lng: -56.5


The image shows a paved two-lane road through extremely flat terrain with tropical/subtropical vegetation consisting of tall grasses and scattered low shrubs/trees. This landscape is characteristic of the Pantanal region in Brazil (Mato Grosso or Mato Grosso do Sul states). The flatness, vegetation type, road marking style, and wooden utility poles with cross-arms are consistent with highways crossing this wetland/savanna region. The clear blue sky and lush green vegetation suggest a wet season in a tropical climate. Similar roads include the BR-262 or MS-418 which traverse the Pantanal, though the exact road cannot be determined without additional landmarks.


Methodology Note

Each model receives the same Google Street View image. Models are tasked with guessing the location of the image using only visual clues from the street view. Their answers are evaluated without feedback or retries.

Leaderboard