PatientAgentBench - Task Completion: leaderboard
The task completion rubric of PatientAgentBench (2026): the mean 1 to 5 jury score for resolving the primary intent of the simulated patient, with top scores requiring execution through the correct tool workflow rather than acknowledgment, across 1,200 conversations. Higher is better.
Metric: Mean task completion rubric score (1-5); 1,200 synthetic patient conversations of up to 15 turns with a simulated patient, the model inside the benchmark's baseline ReAct agent harness with 15 healthcare sandbox tools; jury of Claude Opus 4.8 and GPT-5.5 judges scoring clinician-grounded rubrics on a 1 to 5 scale; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 5 | 4.86 | #111 |
| 2 | GPT-5.4 | 4.85 | #123 |
| 3 | Claude Opus 4.8 | 4.84 | #56 |
| 4 | GPT-5.5 | 4.79 | #49 |
| 5 | Gemini 3 Flash | 4.69 | #164 |
| 6 | Gemini 3.1 Pro (Preview) | 4.65 | #94 |
| 7 | GPT-OSS-120B | 4.56 | #700 |
| 8 | Claude Haiku 4.5 | 4.52 | #568 |
| 9 | Qwen 3 235B A22B | 4.49 | #623 |
| 10 | Qwen3-Next-80B | 4.17 | #626 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=patientagentbench-task-completion · How It Works · Data refreshed daily, snapshot 2026-11-13.