PanCanBench - Synthetic Rubrics: leaderboard
Figure5 reproduction: same6203eligible saved responses across22models, ClaudeOpus5 judge, final human3639-item/282question rubric and separate synthetic rubric. Mean per-question percentage includes negative penalties and genuinezeros, excludes explicit NO RESPONSE GENERATED. Synthetic35interpretation/direction flags remain unadjudicated; no manual score changes. Not pooled-points percentage or the earlier3130-item manuscript draft. Synthetic Rubrics. Higher is better; negative percentages permitted by rubric penalties.
Metric: Mean rubric score (%). Source: github.com. Saturation forecast: Not forecast. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 96.02 | #307 |
| 2 | O3 | 94.4 | #228 |
| 3 | GPT-5 | 93.17 | #167 |
| 4 | Gemini 2.5 Flash | 92.21 | #453 |
| 5 | Gemini 2.5 Pro | 89.58 | #266 |
| 6 | O4 Mini | 89.45 | #353 |
| 7 | OLMo 3 32B (Thinking) | 87.32 | #1884 (OLMo 3 32B) |
| 8 | Gemma 3 27B (IT) | 86.79 | #1289 |
| 9 | GPT-4.1 | 86.62 | #495 |
| 10 | Gemma 3 12B (IT) | 86.21 | #1837 |
| 11 | Qwen 3 32B | 83.37 | #960 |
| 12 | OLMo 3.1 32B Instruct | 81.14 | #2384 |
| 13 | GPT-4o | 78.3 | #741 |
| 14 | Qwen 3 8B | 77.11 | #1830 |
| 15 | Qwen 3 14B | 76.84 | #1233 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=pancanbench-synthetic-rubrics · How It Works · Data refreshed daily, snapshot 2026-11-13.