GeoBench ACW: leaderboard
GeoGuessr for LLMs (ACW subset) - identify location from a single static image using the Around-the-Clock-and-World dataset. Tests visual geographic reasoning and world knowledge.
Metric: Average Score. Source: geobench.org. Saturation forecast: Around January 2027. 54 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 | 4403.7 | #16 |
| 2 | Gemini 3 Flash (Preview) | 4332.77 | #82 |
| 3 | Gemini 3.8 Flash | 4306.9 | #21 |
| 4 | Claude Fable 5.1 | 4302.89 | #9 |
| 5 | Gemini 3.5 Flash | 4288.21 | #60 |
| 6 | Gemini 3.1 Flash Lite (Preview) | 4229.08 | #194 |
| 7 | Gemini 3.1 Pro (Preview) | 4172.19 | #58 |
| 8 | Gemini 3 Pro (Preview) | 4144.58 | #66 |
| 9 | Claude Opus 4.8 (Thinking) | 4127.75 | #37 (Claude Opus 4.8) |
| 10 | Muse Spark 1.3 | 4102.77 | #24 |
| 11 | Gemini 2.5 Pro (Preview 03-25) | 4092.5 | #123 |
| 12 | Gemini 2.5 Pro (Preview 05-06) | 4085.44 | #147 |
| 13 | Gemini 2.5 Pro | 4019.35 | #152 |
| 14 | GPT-5 | 3937 | #98 |
| 15 | Gemini 1.5 Flash | 3933.94 | #618 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=geobench-acw · How It Works · Data refreshed daily, snapshot 2026-10-20.