BenchmarksHealthcare · RxSafeBench
HealthcareAcademic benchmark

RxSafeBench

Updated: 7/22/2026

Tests whether a medical agent can prescribe safely when drugs interact - picking the medication that treats the indication without harming the patient's existing regimen.

1,380 cases·5 systems

Pramaana scores 100% across everything

Every interaction case, every indication - no dangerous combination slips through Pramaana's formalized prescription graph.

Cases
1,380
Indications
696
HEALTHCARE AGENT VERIFIED · PRAMAANA LABS100%VERIFIED

Results

#SystemCase pass rateSolved
1
PramaanaPramaana Labs
100.0% ± 0.01,380/1,380
2
Claude Fable 5 (max)AnthropicBest frontier model
54.6% ± 1.3754/1,380
3
GPT-5.6 Sol (xhigh)OpenAI
52.3% ± 1.3722/1,380
4
Muse Spark 1.1 (xhigh)Meta
50.2% ± 1.3693/1,380
5
Qwen3.6 27B (thinking)Qwen
44.1% ± 1.3608/1,380

Pass rates are measured over 1,380 verified drug-interaction cases with all-or-nothing grading per case. Muse Spark 1.1 returned errors on 4 cases; errored cases are counted as failures.

Overview

When multiple drugs are prescribed to the same patient, the drugs can interact in ways that are harmful - blood thinners combined with common painkillers like ibuprofen raise the chance of internal bleeding. A critical safety test for medical agents is whether they can prescribe under these combinatorial constraints.

RxSafeBench is the verified drug-interaction subset of RxSafeBench (IEEE BIBM 2025), a simulated-consultation benchmark built on a curated risk database of 28,781 drug-drug interactions. Its 1,380 interaction cases span 696 indications; each presents a patient already taking a risk drug, and the agent must select the medication that treats the indication without triggering a dangerous interaction.

We formalized the prescription relationship between drugs. Pramaana's formalized interaction graph achieves 100% while frontier LLM guesswork remains dangerously inaccurate.

Open the RxSafeBench paper

Key takeaways

  • Frontier models routinely green-light dangerous prescription combinations. The best frontier model - Claude Fable 5 with max thinking - passes 54.6% of cases. LLMs reach for the most common prescription for the illness without checking how it interacts with the rest of the regimen.
  • Models get the same scenarios wrong, so going multi-model won't help. The three strongest frontier models fail on heavily overlapping clinical scenarios - pairwise Jaccard similarity between their failure sets runs 0.66 to 0.72. Ensembling frontier models is unlikely to remove the danger.

Benchmark scope

  • Read a simulated consultation where the patient already takes a risk drug
  • Choose the drug that treats the indication from options that include dangerous interactions
  • Avoid therapeutic duplication and contraindicated drug-drug combinations
  • All-or-nothing grading per case - only the safe, indicated drug passes

Updates

RxSafeBench

Launch set: 1,380 verified interaction cases across 696 indications, evaluated on four frontier models against Pramaana's formalized prescription baseline. Pramaana passes every case.