Harbor Adapters - MedAgentBench (Terminus-2): leaderboard
Metric: Success rate (%; 100 tasks, mean of 3 trials). Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 63.7 |
| 2 | Claude Opus 4.6 | 59.7 |
| 3 | Gemini 3 Flash (Preview) | 50.7 |
| 4 | Claude Sonnet 4.6 | 49.7 |
| 5 | Claude Haiku 4.5 | 46.7 |
| 6 | GPT-5 Mini | 35 |
| 7 | GPT-5.4 (Medium) | 34 |
| 8 | GPT-5 Nano | 8 |
Interactive version: theaggregate.ai/benchmark?slug=harbor-adapters-medagentbench-terminus-2 · How It Works · Data refreshed daily, snapshot 2026-09-25.