GeoCode Leaderboard: leaderboard
Leaderboard for large language models on geospatial code generation using AutoGEEval and AutoGEEval++ pass@k metrics on Google Earth Engine tasks.
Metric: AutoGEEval++ Pass@1 (imported). Source: szx-0633.github.io. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 70.93 | #248 |
| 2 | Claude 3.7 Sonnet | 70.35 | #249 |
| 3 | DeepSeek V3 (0324) | 70.25 | #340 |
| 4 | Qwen 2.5 Coder 32B | 66.77 | #667 |
| 5 | GPT-4.1 Mini | 66.56 | #355 |
| 6 | Qwen 2.5 32B | 65.45 | #546 |
| 7 | QwQ-32B | 60.69 | #421 |
| 8 | Qwen 3 32B | 60.67 | #443 |
| 9 | DeepSeek R1 | 60.42 | #251 |
| 10 | Qwen 2.5 Coder 7B | 50.51 | #1286 |
| 11 | Qwen 2.5 7B | 49.58 | #922 |
| 12 | Qwen 3 8B | 48.35 | #674 |
| 13 | Qwen 3 8B (Thinking) | 38.93 | #674 (Qwen 3 8B) |
| 14 | Qwen 2.5 3B | 36.26 | #1211 |
| 15 | Qwen 3 4B | 34.03 | #829 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=geocode-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.