SpatialBench-Long V1 Verified
Science
SpatialBench-Long V1 verified 22-task subset of the 24-task long-horizon spatial-transcriptomics release. Published per-model and per-harness pass rate.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Terra | 37.9 |
| GPT-5.6 Sol | 36.4 |
| DeepSeek V4.1 Flash | 34.9 |
| GPT-6 Astra | 34.9 |
| GPT-5.5 | 33.3 |
| Claude Opus 5 | 31.8 |
| Gemini 3.5 Flash | 30.3 |
| Grok 4.6 | 30.3 |
| Opus 4.8 | 28.8 |
| Kimi K3 | 28.8 |
| GPT-5.6 Luna | 27.3 |
| Sonnet 5 | 24.2 |
| Grok 4.5 | 22.7 |