GeoBench
Rank #19 of 51 · 26 models scored · top score 88
unknown
Strongest on discrimination, weakest on contamination resistance. Caveat: the public test set leaves contamination on the table.
Weighted composite. Each signal is computed from the source-backed leaderboard below.
Does it still separate models?
How far from ceiling / clustered at the top?
Public vs held-out; training-leak risk.
Apples-to-apples, or mixed / vendor-optimized?
How old is the benchmark?
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | Gemini 3 Flash | 88 | unverified | 2025-12-17 |
| 2 | Gemini 2.5 Pro (May 2025) | 86 | unverified | 2025-05-06 |
| 3 | Gemini 3 Pro | 84 | unverified | 2025-11-18 |
| 4 | GPT-5 | 81 | unverified | 2025-08-07 |
| 5 | Gemini 2.5 Pro (Mar 2025) | 81 | unverified | 2025-03-25 |
| 6 | o1 | 80 | unverified | 2024-12-05 |
| 7 | Gemini 2.0 Flash (Feb 2025) | 77 | unverified | 2024-12-11 |
| 8 | Gemini 1.5 Flash (Sep 2024) | 76 | unverified | 2024-05-10 |
| 9 | Gemini 2.5 Flash (May 2025) | 76 | unverified | 2025-05-20 |
| 10 | Claude Opus 4.5 | 75 | unverified | 2025-11-24 |
| 11 | o3 | 74 | unverified | 2024-12-20 |
| 12 | Gemini 2.5 Flash (Apr 2025) | 73 | unverified | 2025-04-17 |
| 13 | GPT-4.1 | 72 | unverified | 2025-04-14 |
| 14 | GPT-4o (Nov 2024) | 71 | unverified | 2024-05-13 |
| 15 | Claude 3.7 Sonnet | 68 | unverified | 2025-02-24 |
| 16 | o4-mini | 64 | unverified | 2025-04-16 |
| 17 | GPT-4o mini | 64 | unverified | 2024-07-18 |
| 18 | Claude 3.5 Sonnet (October 2024) | 62 | unverified | 2024-10-22 |
| 19 | Qwen2.5-72B | 62 | unverified | 2024-09-19 |
| 20 | Gemma 3 27B | 52 | unverified | 2025-03-12 |
| 21 | Llama 3.2 90B | 52 | unverified | 2024-09-24 |
| 22 | Llama 4 Maverick | 52 | unverified | 2025-04-05 |
| 23 | Claude Opus 4 | 49 | unverified | 2025-05-22 |
| 24 | Grok 4 | 45 | unverified | 2025-07-09 |
| 25 | Claude Sonnet 4 | 37 | unverified | 2025-05-22 |
| 26 | Claude 3.5 Haiku | 34 | unverified | 2024-10-22 |