Objective:

To evaluate the performance of DeepSeek (R1 version) in clinical scenarios and compare it with GPT-5.3.

Approach:

Key Findings:

• DeepSeek-R1 achieved 94.9% appropriateness and 100% helpfulness.
• 91.7% of DeepSeek-R1's responses to follow-up inquiries were rated empathetic.
• 7.7% of DeepSeek-R1's responses showed substantial inconsistencies, primarily in tumor staging.
• GPT-5.3 exhibited equivalent core performance with 94.9% appropriateness and 100% helpfulness but lower empathy (66.7%).
• Both models had similar primary diagnosis accuracy (10%) and differential diagnosis accuracy (60%).

Interpretation:

Limitations:

• DeepSeek-R1 had a 7.7% inconsistency rate in responses.
• Only 37% of DeepSeek-R1's cited references were fully valid.
• Both models showed limitations in diagnostic accuracy.

Conclusion:

Attribution Notice

This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.