GeoBuildBench - Level 3 (Medium): leaderboard
Metric: Construction success rate (%; the agent's DSL program must create every required object and satisfy every verification condition within 5 interaction steps; 162 level-3 problems, rate = printed success count / 162). Source: arxiv.org. Saturation forecast: Around December 2026. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.1 | 76.54 |
| 2 | Gemini 3 Flash | 72.22 |
| 3 | Qwen 3 VL 235B A22B Instruct | 34.57 |
| 4 | Llama 3.2 90B Vision Instruct | 13.58 |
Interactive version: theaggregate.ai/benchmark?slug=geobuildbench-level-3-medium · How It Works · Data refreshed daily, snapshot 2026-09-26.