GPSBench (Pure GPS) - Relative Position: leaderboard
Metric: Score (%) on the relative position (compass direction between two coordinates, usually on different continents) task of GPSBench's Pure GPS Track (geometric computation on coordinates); 1,020 test samples; zero-shot from the model's own knowledge, no tools, no chain-of-thought or few-shot prompting, temperature 0; choice tasks score accuracy and numerical tasks one minus the mean absolute percentage error; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash (Non-reasoning) | 70.8 | #237 (Gemini 2.5 Flash) |
| 2 | GPT-5.1 (Low) | 70.7 | #131 (GPT-5.1) |
| 3 | Gemini 2.5 Pro | 69.7 | #145 |
| 4 | GPT-5 Nano (Low) | 65.5 | #415 (GPT-5 Nano) |
| 5 | Qwen 3 30B A3B 2507 Instruct | 64.1 | #464 |
| 6 | Claude Haiku 4.5 | 58.1 | #271 |
| 7 | Mistral Large 3 | 57.6 | #388 |
| 8 | GPT-4.1 Mini | 51 | #346 |
| 9 | GPT-4.1 | 47.5 | #240 |
| 10 | GPT-5 Mini (Low) | 46.6 | #176 (GPT-5 Mini) |
| 11 | Qwen 3 235B A22B 2507 Instruct | 42.5 | #291 |
| 12 | Mistral Small 3 | 38.6 | #616 |
| 13 | Qwen 3 14B (Non-reasoning) | 35.8 | #524 (Qwen 3 14B) |
| 14 | Qwen 3 8B (Non-reasoning) | 32.4 | #667 (Qwen 3 8B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=gpsbench-pure-gps-relative-position · How It Works · Data refreshed daily, snapshot 2026-10-11.