Geoguessr
Released: 2026-03-07
Key:
Score Threshold:
google/gemini-3-flash-preview
Score: 4654
Distance: 143km
Lat: -13.0645069
Lng: -55.9324683
The image shows a large-scale agricultural landscape with red soil, which is highly characteristic of the Brazilian ’Cerrado’ region, specifically in states like Mato Grosso. The flat horizon, vast fields, power lines, and cloudy sky are common features along major highways such as the BR-163, which serves as a primary transport corridor for soy and corn production in central Brazil.
meta-llama/llama-4-maverick
meta
Score: 2447
Distance: 1429km
Lat: -23.5532
Lng: -46.6623
The image shows a vast, open field with red soil and sparse vegetation, indicating a rural or countryside area. The presence of power transmission lines in the distance suggests it is near an infrastructure development. The overall landscape and cloud cover are consistent with regions that have a subtropical or tropical climate. Based on these features and the fact that Google Street View is more comprehensive in certain countries, I guessed a location in Brazil, which is known for its diverse landscapes and widespread use of Google Street View.
anthropic/claude-sonnet-4.6
anthropic
Score: 3275
Distance: 847km
Lat: -18.5
Lng: -49.5
The image shows red laterite soil, flat agricultural land with green crops (likely sugarcane or soybeans), power transmission lines, and overcast tropical sky. The distinctive red/orange soil color is characteristic of the Brazilian Cerrado region, particularly in the states of Goiás or Minas Gerais/São Paulo area. The flat terrain, extensive agriculture, and soil color strongly suggest the interior of Brazil, likely in Goiás or western Minas Gerais.
anthropic/claude-haiku-4.5
anthropic
Score: 17
Distance: 11392km
Lat: 50.2
Lng: 36.2
This image shows a vast, flat agricultural landscape with reddish-brown soil typical of the steppe regions of central Russia. The flat terrain, the style of utility poles with multiple power lines, the vegetation pattern, and the distinctive reddish earth are characteristic of the Voronezh Oblast or Tambov Oblast region in southwestern Russia. The agricultural development and the specific soil color point to this area of the Russian steppe. The image appears to be taken along a rural road in this region.
openai/gpt-5-image-mini
openai
Score: 3318
Distance: 820km
Lat: -17
Lng: -49
Flat, extensive agricultural plain with very red tropical clay soil (oxidized ’terra roxa’ / Oxisol), sparse low vegetation characteristic of Brazil’s Cerrado. Lattice high‑voltage transmission towers and the style of roadside drainage are typical of inland Brazil rather than Australia or Africa. Cloudy tropical sky and large-scale row cropping suggest central Brazil (Goiás/Mato Grosso do Sul region). Coordinates placed in central Goiás as a best estimate.
x-ai/grok-4-fast
x-ai
Score: 236
Distance: 6105km
Lat: 34
Lng: -83.5
The image depicts a rural landscape with reddish soil, characteristic of laterite or clay-rich soils common in the southeastern United States, such as Georgia. The flat green fields and power lines suggest a countryside road, likely in the Piedmont region. This guess is based on the soil color and vegetation patterns visible in the Street View imagery.
moonshotai/kimi-k2.5
moonshot
Score: 3
Distance: 15020km
Lat: 13.5
Lng: 77.5
The image shows distinctive red/orange laterite soil, which is characteristic of the Deccan Plateau region in India, particularly in states like Karnataka, Andhra Pradesh, and Telangana. The flat agricultural landscape with green fields in the background, combined with the presence of high-voltage transmission towers, is typical of rural areas in this region. The overcast sky and the type of vegetation debris in the foreground are consistent with the monsoon climate of peninsular India. Based on the soil type and landscape features, this appears to be in the interior regions of Karnataka, approximately between Bangalore and Tumkur.
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.