MedCUA-Bench - Intent Goals: leaderboard
Metric: Strict success rate (%) on the 216 intent-goal tasks (the clinical outcome a senior clinician would dictate) of 432 screenshot-only clinical computer-use tasks (18 scenarios in 10 medical domains: 15 synthetic reconstructions of clinical software, the real OpenEMR EHR and the OHIF imaging viewer; 216 base tasks each under an intent goal and a step goal), one run per task in the shared AgentLab harness with a 30-step budget; strict success requires task completion with no safety violation; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 51.9 |
| 2 | Claude Opus 4.7 | 50.9 |
| 3 | Claude Sonnet 4.6 | 43.1 |
| 4 | Kimi K2.6 | 28.2 |
| 5 | Qwen 2.5 VL 32B Instruct | 18.5 |
| 6 | GPT-5 Mini | 14.4 |
| 7 | Gemini 3.1 Pro (Preview) | 5.1 |
| 8 | Gemma 3 27B | 4.2 |
| 9 | InternVL3.5-8B | 3.2 |
| 10 | Qwen 3.5 27B | 2.3 |
| 11 | Qwen 3.5 9B | 1.4 |
| 12 | Gemma 4 E4B | 0.9 |
Interactive version: theaggregate.ai/benchmark?slug=medcua-bench-intent-goals · How It Works · Data refreshed daily, snapshot 2026-09-29.