Atlas

Benchmarks

← All benchmarks

HardGeoBench Country Accuracy

Multimodal · 2025-05-08

HardGeoBench country-level accuracy for private-photo geolocation predictions. Scores are pass@1 percentages from API default settings with a simple prompt, no sampling, and no web search or tools.

Top models (higher is better)

ModelScore
Gemini 3.5 Flash66.3
Claude Fable 561.3
Gemini 3 Flash Preview60.0
Gemini 3.1 Pro Preview60.0
Gemini 3 Pro Preview58.8
Gemini 2.5 Pro Experimental 03-2556.3
Gemini 2.5 Pro Preview 06-0556.3
Sonnet 4.656.3
Gemini 2.0 Flash-Lite55.0
GPT-5 (2025-08-07)55.0
Opus 4.655.0
o3-pro53.8
Gemini 2.5 Flash Preview (09-2025)53.8
Gemini 3.1 Flash-Lite Preview53.8
Opus 4.753.8
Opus 4.853.8
Gemini 2.5 Flash Preview 05-2052.5
ChatGPT-4o Latest (source-unspecified snapshot)52.5
GPT-4.151.3
o351.3
Gemini 2.0 Flash51.3
GPT-550.0
GPT-5.150.0
Opus 4.550.0
Gemini 1.5 Flash48.8
Sonnet 547.5
GPT-5.6 Sol47.5
GPT-5.2 (2025-12-11)45.0
Kimi K2.545.0
GPT-5.545.0
GPT-5.6 Terra45.0
GPT-5 Mini43.8
GPT-5.442.5
O4 Mini41.3
Haiku 3.540.0
Grok 4.1 Fast40.0
Gemma 4 31B IT40.0
Grok 4.540.0
Sonnet 4.537.5
Grok 4.2037.5
Loading Atlas data…