GPSBench (Applied) - Route Analysis: leaderboard
Metric: Score (%) on the route analysis (whether a named waypoint city lies on the route between two named cities) task of GPSBench's Applied Track (geographic reasoning that combines coordinates or place names with world knowledge); 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: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 91.7 | #145 |
| 2 | GPT-5 Mini (Low) | 90.1 | #176 (GPT-5 Mini) |
| 3 | Claude Haiku 4.5 | 89.7 | #271 |
| 4 | GPT-4.1 Mini | 89.3 | #346 |
| 5 | GPT-4.1 | 88.6 | #240 |
| 6 | Qwen 3 235B A22B 2507 Instruct | 88.3 | #291 |
| 7 | Mistral Large 3 | 87.9 | #388 |
| 8 | Gemini 2.5 Flash (Non-reasoning) | 86.9 | #237 (Gemini 2.5 Flash) |
| 9 | Qwen 3 30B A3B 2507 Instruct | 85.7 | #464 |
| 10 | GPT-5 Nano (Low) | 84.8 | #415 (GPT-5 Nano) |
| 11 | Qwen 3 14B (Non-reasoning) | 83.2 | #524 (Qwen 3 14B) |
| 12 | Mistral Small 3 | 82 | #616 |
| 13 | GPT-5.1 (Low) | 81.7 | #131 (GPT-5.1) |
| 14 | Qwen 3 8B (Non-reasoning) | 80.1 | #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-applied-route-analysis · How It Works · Data refreshed daily, snapshot 2026-10-11.