FHIR-AgentBench (Single-Turn + Retriever + Code): leaderboard
Metric: Answer correctness (%; o4-mini judge of the free-text answer against the SQL-derived answer, 2,931 questions; one turn with a retriever and a code interpreter). Source: arxiv.org. Saturation forecast: Around 2029. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Flash | 39 |
| 2 | Llama 3.3 70B | 36 |
| 3 | Qwen 3 32B | 33 |
| 4 | O4 Mini | 33 |
Interactive version: theaggregate.ai/benchmark?slug=fhir-agentbench-single-turn-plus-retriever-plus-code · How It Works · Data refreshed daily, snapshot 2026-09-26.