Updated daily at 05:00 UTC
Benchmark
GeoBench
Measures: GeoGuessr-style geolocation: from a street-level photo the model gives coordinates and a country; the headline is the share of countries named correctly.
- Maintainer
- ccmdi
- Introduced
- 2025
- Status
- Active
- Saturation
- Near-ceiling
- Typical provenance
- Independent
- In scoring set
- Yes
Leaderboard
Independent and third-party results only.
| Model | Country accuracy | Provenance | Source | Evaluated |
|---|---|---|---|---|
| Gemini 3 Flash | 88.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 2.5 Pro | 86.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 3 Pro | 84.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 2.5 Pro | 81.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 2.5 Pro | 81.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| GPT-5 OpenAI · Medium effort | 81.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| o1 OpenAI · Medium effort | 80.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 2.0 Flash | 77.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 1.5 Flash | 76.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 2.5 Flash | 76.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude Opus 4.5 | 75.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| o3 OpenAI · Medium effort | 74.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemini 2.5 Flash | 73.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| GPT-4.1 | 72.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| GPT-4o | 71.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude 3.7 Sonnet | 68.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude 3.7 Sonnet Anthropic · 15K effort | 65.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| GPT-4o mini | 64.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| o4-mini OpenAI · High effort | 64.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| o4-mini OpenAI · Medium effort | 64.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude 3.5 Sonnet | 62.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Qwen2.5-VL-72B-Instruct | 62.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| o3 OpenAI · High effort | 60.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Gemma 3 27B | 52.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Llama 3.2 90B Vision | 52.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Llama 4 Maverick | 52.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude Opus 4 Anthropic · 32K effort | 49.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Grok 4 | 45.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude Sonnet 4 | 37.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Pixtral 12B | 37.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude 3.5 Haiku | 34.0% | Third party | GeoBench leaderboard via Epoch AI | — |
| Claude Sonnet 4 Anthropic · 32K effort | 32.0% | Third party | GeoBench leaderboard via Epoch AI | — |
- Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks, CC BY 4.0.