IBA-Bench - Medical Services: leaderboard
Metric: Checkpoint satisfaction (%; share of hidden preference checkpoints satisfied by the final action, medical services domain; mean of three LLM judges: DeepSeek-V3.2, Qwen3-4B-Instruct and Qwen3-30B-A3B-Instruct; retrieval-augmented baseline with bge-m3 top-3 chunks). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 4B Instruct | 82.9 |
| 2 | GPT-5 Mini | 80.2 |
| 3 | QwQ-32B | 79.6 |
| 4 | DeepSeek V3.2 | 79 |
| 5 | DeepSeek V3 | 78.8 |
| 6 | GPT-5.1 | 78 |
| 7 | Qwen 2.5 7B Instruct | 75 |
| 8 | GPT-4o | 73 |
| 9 | GPT-4o Mini | 71.8 |
| 10 | GLM-4 9B Chat | 69.3 |
Interactive version: theaggregate.ai/benchmark?slug=iba-bench-medical-services · How It Works · Data refreshed daily, snapshot 2026-09-26.