GeoBench Photos: leaderboard
GeoGuessr for LLMs (Photos subset) - identify location from a single photograph. Tests visual geographic reasoning and world knowledge from one photo.
Metric: Average Score. Source: geobench.org. Saturation forecast: Around January 2027. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 | 4313.64 | #16 |
| 2 | Claude Fable 5.1 | 4259.62 | #9 |
| 3 | Gemini 3.8 Flash | 4251.18 | #21 |
| 4 | Gemini 3 Flash (Preview) | 4204.26 | #82 |
| 5 | Muse Spark 1.3 | 4202.16 | #24 |
| 6 | Claude Opus 4.8 (Thinking) | 4101.62 | #37 (Claude Opus 4.8) |
| 7 | Gemini 3.1 Flash Lite (Preview) | 3964.1 | #194 |
| 8 | Gemini 3 Pro (Preview) | 3893.24 | #66 |
| 9 | GPT-5.6 Sol (High) | 3881.16 | #19 (GPT-5.6 Sol) |
| 10 | Gemini 2.5 Flash (Preview 04-17) | 3873.02 | #237 |
| 11 | Gemini 2.5 Pro (Preview 03-25) | 3870.6 | #123 |
| 12 | Gemini 3.5 Flash | 3844.88 | #60 |
| 13 | Gemini 2.5 Pro (Preview 05-06) | 3835.62 | #147 |
| 14 | Claude Opus 4.5 (Thinking) | 3796.96 | #83 (Claude Opus 4.5) |
| 15 | O3 (High) | 3788.53 | #124 (O3) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=geobench-photos · How It Works · Data refreshed daily, snapshot 2026-10-20.