BIM-Edit - Geometry: leaderboard
BIM-Edit (2026) has LLM agents edit Building Information Models in IFC format from natural-language instructions (create, update and delete; direct, spatial or topological references) over 324 tasks. This board is the geometry score on a 0 to 100 scale, how closely the edited geometry matches the reference model; higher is better.
Metric: Geometry score (0-100): exponential decay of the median bidirectional Chamfer distance between surface point clouds of the edited and the reference elements, on 324 natural-language create, update and delete tasks on 11 real and 36 synthetic IFC building models (direct, spatial and topological instructions), the model edits the IFC file as an agent that runs Python through a single IfcOpenShell execution tool with a 20-call budget, temperature 0 where settable; mean over tasks; higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 68.87 | #144 |
| 2 | Qwen 3.6 Plus | 59.21 | #222 |
| 3 | Claude Sonnet 4.6 | 55.9 | #161 |
| 4 | DeepSeek V3.2 | 50.97 | #406 |
| 5 | GPT-5.4 Pro (xHigh) | 50.14 | #85 (GPT-5.4 Pro) |
| 6 | GPT-5.4 Mini | 45.18 | #389 |
| 7 | Gemma 4 31B (IT) | 44.94 | #342 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=bim-edit-geometry · How It Works · Data refreshed daily, snapshot 2026-11-13.