GeoBenchLLM - GridRoute: leaderboard
Metric: Optimal route ratio (%). Source: arxiv.org. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 8B (Thinking) | 81 |
| 2 | GPT-OSS-120B (Low) | 80 |
| 3 | GPT-OSS-20B (Low) | 67 |
| 4 | Qwen 3 1.7B (Thinking) | 66 |
| 5 | Qwen 3 8B (Non-reasoning) | 51 |
| 6 | Qwen 3 0.6B (Thinking) | 41 |
Interactive version: theaggregate.ai/benchmark?slug=geobenchllm-gridroute · How It Works · Data refreshed daily, snapshot 2026-09-19.