Benchmarks
Geobench
Models scored
32
Best score
4333
Metric
ACW Avg Score
Geobench is an AI evaluation tracked by Epoch AI, with 32 scored model versions on record. The scores here are reported by third parties rather than produced by Epoch AI running the evaluation itself.
Scored models span July 2024 to December 2025. The best score recorded is 4333, measured as ACW Avg Score. Models come from Google DeepMind, OpenAI, Google DeepMind,Google, Anthropic, Alibaba, Meta AI, xAI, Mistral AI.
Full record
- Score metric
- ACW Avg Score
- Models scored
- 32
- Best score recorded
- 4333
- Earliest model scored
- Jul 18, 2024
- Most recent model scored
- Dec 17, 2025
- Third-party reported
- Yes
Leaderboard
| 01 | gemini-3-flash-preview | Google DeepMind | 4333 |
| 02 | gemini-3-pro-preview | Google DeepMind | 4145 |
| 03 | gemini-2.5-pro-preview-05-06 | Google DeepMind | 4085 |
| 04 | gemini-2.5-pro-exp-03-25 | Google DeepMind | 4019 |
| 05 | gpt-5-2025-08-07_medium | OpenAI | 3937 |
| 06 | gemini-1.5-flash-002 | Google DeepMind | 3934 |
| 07 | o3-2025-04-16_high | OpenAI | 3930 |
| 08 | o1-2024-12-17_medium | OpenAI | 3925 |
| 09 | gemini-2.0-flash-001 | Google DeepMind,Google | 3897 |
| 10 | claude-opus-4-5-20251101 | Anthropic | 3877 |
| 11 | o3-2025-04-16_medium | OpenAI | 3858 |
| 12 | gemini-2.5-flash-preview-04-17 | Google DeepMind | 3854 |
| 13 | claude-3-7-sonnet-20250219_15K | Anthropic | 3844 |
| 14 | gpt-4.1-2025-04-14 | OpenAI | 3791 |
| 15 | claude-3-7-sonnet-20250219 | Anthropic | 3791 |
| 16 | gemini-2.5-flash-preview-05-20 | Google DeepMind | 3786 |
| 17 | o4-mini-2025-04-16_high | OpenAI | 3717 |
| 18 | gpt-4o-2024-11-20 | OpenAI | 3696 |
| 19 | o4-mini-2025-04-16_medium | OpenAI | 3651 |
| 20 | Qwen2.5-VL-72B-Instruct | Alibaba | 3448 |
| 21 | gpt-4o-mini-2024-07-18 | OpenAI | 3401 |
| 22 | claude-3-5-sonnet-20241022 | Anthropic | 3327 |
| 23 | gemma-3-27b-it | Google DeepMind | 3087 |
| 24 | Llama-4-Maverick-17B-128E-Instruct | Meta AI | 3069 |
| 25 | claude-opus-4-20250514_32K | Anthropic | 2915 |
| 26 | Llama-3.2-90B-Vision-Instruct | Meta AI | 2798 |
| 27 | grok-4-0709 | xAI | 2719 |
| 28 | claude-3-5-haiku-20241022 | Anthropic | 2507 |
| 29 | claude-sonnet-4-20250514 | Anthropic | 2468 |
| 30 | claude-sonnet-4-20250514_32K | Anthropic | 2369 |
| 31 | Pixtral-12B-2409 | Mistral AI | 2131 |
More benchmarks
SourceEpoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/benchmarks. Licensed under CC BY 4.0.
← All benchmarks