SpatialBench-Long
Science · 2026-05-27
Long-context spatial transcriptomics benchmark on real tissue datasets, spanning multiple studies and spatial analysis tasks. Scores average three independent runs for each of the 24 distinct tasks; the 72 rollouts are not distinct evaluation items.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.8 | 11.1 |
| Gemini 3.5 Flash | 11.1 |
| GPT-5.5 | 11.1 |
| Opus 4.6 | 9.7 |
| Opus 4.7 | 8.3 |
| Grok 4.20 | 6.9 |
| GPT-5.4 | 5.6 |
| Kimi K2.6 | 5.6 |
| Sonnet 4.6 | 4.2 |
| Gemini 3.1 Pro Preview | 4.2 |
| Grok 4.3 | 4.2 |
| Gemini 2.5 Pro | 1.4 |