# Benchmarking large language models for medical education: performance on the clinical laboratory technician qualification examination

## TL;DR

This study evaluates how well large language models perform on a Chinese clinical laboratory technician exam, showing strong potential for AI in medical education.

## Contribution

The paper introduces a benchmark dataset of 1,600 exam questions and evaluates 12 LLMs for their performance in clinical laboratory technology.

## Key findings

- Qwen3-235B achieved the highest overall accuracy of 89.93% on the CCLTQE exam.
- LLMs optimized for Chinese language and domain-specific content performed exceptionally well.
- Top-performing models included DeepSeek-R1 and QwQ-32B with over 89% accuracy.

## Abstract

Large language models (LLMs) have shown growing applications in medicine, yet their capabilities in the field of clinical laboratory technology remain underexplored. This study aims to evaluate the performance of LLMs in the Chinese Clinical Laboratory Technologist Qualification Examination (CCLTQE) and provide empirical evidence for their application in laboratory medicine. A dataset containing 1,600 single-choice questions is constructed for the CCLTQE exam. The dataset covers four sections: clinical laboratory fundamentals, other medical knowledge related to clinical laboratory technology, clinical laboratory specialized knowledge, and clinical laboratory professional practice competence. We select 12 LLMs for evaluation, including the DeepSeek, GPT, Llama, Qwen, and Gemma series. Results show that Qwen3-235B achieves the highest overall accuracy (89.93%), followed by DeepSeek-R1 (89.75%) and QwQ-32B (89.22%). This study demonstrates that LLMs optimized for Chinese language and domain-specific content demonstrate outstanding performance in CCLTQE, indicating significant potential for AI-assisted education and practice in laboratory medicine.