QSTRBench: leaderboard
QSTRBench evaluates large language models on qualitative spatial and temporal reasoning over calculi such as Point Algebra, Allen’s Interval Algebra, RCC variants, and related relations. The inputs are text. The pinned metric is Overall Accuracy (%), measuring the percentage of questions answered correctly. Source: BenchmarkList’s attributed result table.
Metric: Overall Accuracy (%). Source: benchmarklist.com. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 85.3 | #115 |
| 2 | DeepSeek V3.2 | 72.1 | #210 |
| 3 | Grok 4.1 Fast | 68.4 | #227 |
| 4 | GPT-5.1 | 67.2 | #145 |
| 5 | DeepSeek V3.1 | 65.8 | #275 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qstrbench · How It Works · Data refreshed daily, snapshot 2026-11-05.