ObGynLongBench - Evidence-Only: leaderboard
Metric: Accuracy (%; supporting evidence only). Source: arxiv.org. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 79.1 |
| 2 | DeepSeek V4 Flash | 78.8 |
| 3 | GPT-5.4 Mini | 77.4 |
| 4 | Qwen 3.5 35B A3B (Non-reasoning) | 76.9 |
| 5 | Claude Haiku 4.5 | 75.5 |
| 6 | Qwen 3.5 4B (Non-reasoning) | 75.4 |
| 7 | Qwen 3.5 9B (Non-reasoning) | 73.1 |
| 8 | Qwen 3 VL 4B | 72.1 |
| 9 | Lingshu-7B | 72.1 |
| 10 | Gemma 4 E4B (Non-reasoning) | 68.1 |
| 11 | MedGemma 1.5 4B | 64 |
| 12 | Phi-4 Mini Instruct | 58.7 |
Interactive version: theaggregate.ai/benchmark?slug=obgynlongbench-evidence-only · How It Works · Data refreshed daily, snapshot 2026-09-19.