HardGeoBench Avg Distance
Multimodal · 2025-05-08
HardGeoBench average geodesic distance in miles between a model's predicted photo location and the true location. Lower scores indicate more accurate geolocation; official leaderboard rows use pass@1 raw model calls without web search or tools.
Top models (lower is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 655 |
| Claude Sonnet 5.5 | 683 |
| Gemini 3.1 Pro Preview | 733 |
| Gemini 4 Argon | 741 |
| Gemini 3.6 Flash | 775 |
| Claude Fable 5 | 800 |
| Gemini 3 Flash Preview | 842 |
| Gemini 3.8 Flash | 843 |
| Claude Fable 5.1 | 845 |
| Muse Spark 1.2 | 871 |
| Gemini 3.5 Flash | 886 |
| Muse Spark 1.1 | 889 |
| Gemini 3.7 Flash | 895 |
| GPT-6 Astra | 908 |
| Muse Spark Open | 948 |
| Gemini 3 Pro Preview | 990 |
| Muse Spark 1.3 | 1022 |
| Sonnet 4.6 | 1043 |
| o3 | 1081 |
| Gemini 2.5 Pro | 1119 |
| Gemini 2.0 Flash | 1128 |
| GPT-5 | 1158 |
| Gemini 2.5 Pro Experimental 03-25 | 1172 |
| GPT-5.5 | 1196 |
| Gemini 2.0 Flash-Lite | 1216 |
| Claude Opus 5 | 1218 |
| GPT-6.1 Sol | 1239 |
| Opus 4.8 | 1245 |
| Opus 4.7 | 1279 |
| GPT-5.1 | 1303 |
| GLM-5.3 Flash | 1309 |
| Claude Haiku 5.5 | 1319 |
| o3-pro | 1340 |
| Gemini 2.5 Flash Preview (09-2025) | 1343 |
| Muse Wonder | 1369 |
| GPT-5.6 Sol | 1381 |
| Gemini 3.1 Flash-Lite Preview | 1435 |
| Gemini 2.5 Flash | 1444 |
| GPT-4.1 | 1494 |
| GPT-6 Sol | 1497 |