SceneActBench - Layout: leaderboard
Layout task of SceneActBench (2026): 100 3D-FRONT rooms from one view: place N canonical object meshes in their poses, scored by ADD-S error against a 4 m reference. The agent acts through a shared Blender tool interface; task score from 0 to 100, higher is better.
Metric: Task score (0-100): 100 3D-FRONT rooms from one view: place N canonical object meshes in their poses, scored by ADD-S error against a 4 m reference; every configuration drives one headless Blender through the same MCP tool interface (scene and object inspection, Python execution, rendering) with task step budgets; each native geometric error or F-score is mapped to a 0 to 100 case score against a fixed reference, invalid outputs score 0, cases are averaged per task. Source: arxiv.org. Saturation forecast: Around October 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (High) | 84.1 | #123 (GPT-5.4) |
| 2 | Step 3.7 Flash (High) | 77.5 | #465 (Step 3.7 Flash) |
| 3 | Seed 2.0 Pro (High) | 77.4 | #211 (Seed 2.0 Pro) |
| 4 | Claude Opus 4.6 (High) | 73.5 | #102 (Claude Opus 4.6) |
| 5 | GPT-5.4 (Medium) | 72.7 | #123 (GPT-5.4) |
| 6 | Kimi K2.6 (Thinking) | 70.9 | #178 (Kimi K2.6) |
| 7 | Gemini 3.1 Pro (Preview) (High) | 65.4 | #94 (Gemini 3.1 Pro (Preview)) |
| 8 | MiniMax-M3 (High) | 58.1 | #269 (MiniMax-M3) |
| 9 | Claude Sonnet 5 (High) | 51.9 | #111 (Claude Sonnet 5) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sceneactbench-layout · How It Works · Data refreshed daily, snapshot 2026-11-13.