SpatialBench V1 Clean Reruns
Science
Clean-v1 reruns on 115 spatial-transcriptomics evaluations. Formerly published as V2, now V1; updated evaluation sets remain separate from the original cohort.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 72.8 |
| Claude Sonnet 5.5 | 72.8 |
| Grok 4.7 | 69.6 |
| Claude Opus 5 | 68.4 |
| GPT-5.6 Sol | 68.1 |
| GPT-6 Astra | 67.5 |
| GPT-6 Sol | 67.5 |
| GPT-6.1 Sol | 67.3 |
| Grok 4.6 | 67.3 |
| Opus 4.8 | 67.0 |
| Sonnet 5 | 66.7 |
| GPT-5.5 | 64.1 |
| Kimi K3 | 63.5 |
| DeepSeek V4.1 Flash | 63.2 |
| GPT-5.6 Luna | 60.9 |
| GPT-6 Luna | 59.7 |
| Gemini 3.8 Flash | 59.4 |
| Gemini 3.7 Flash | 59.1 |
| Opus 4.7 | 56.5 |
| Sonnet 4.6 | 53.9 |
| Nemotron 3.5 Lightning 30B A3B (precision unspecified) | 29.0 |