SpatialText - Logical Dynamics: leaderboard
Metric: Accuracy (%) on level V logical dynamics (path planning, counterfactuals, consistency checks, functional inference, 95 questions) on SpatialText's 485 questions about human-annotated text descriptions of real indoor scenes (LSUN), text only, greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3.2 | 76.84 | #198 |
| 2 | Qwen 3 8B | 65.26 | #667 |
| 3 | Gemma 3 12B (IT) | 60 | #655 |
| 4 | Gemma 2 9B (IT) | 54.74 | #774 |
| 5 | DeepSeek R1 Distill Llama 8B | 42.11 | #1282 |
| 6 | Qwen 2.5 7B Instruct | 41.05 | #846 |
| 7 | Mistral Instruct v0.1 7B | 35.79 | #1429 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=spatialtext-logical-dynamics · How It Works · Data refreshed daily, snapshot 2026-10-11.