Dietary guidance for pregnant women using DeepSeek-R1 and ChatGPT-4.0: a comparative analysis

TL;DR

This study compares two AI models, DeepSeek-R1 and ChatGPT-4.0, in generating dietary plans for pregnant women, finding differences in quality and cost.

Contribution

The study evaluates and compares the performance of two AI models in generating dietary guidance for pregnant women.

Findings

DeepSeek-R1 provided better caloric targets and food adequacy scores compared to ChatGPT-4.0.

ChatGPT-4.0 performed better in moderation and cost-effectiveness aspects of dietary plans.

Both models achieved satisfactory DQI-I scores for dietary quality.

Abstract

Advancements in artificial intelligence (AI) and natural language processing have enabled the widespread application of large language models. However, the ability of AI models to provide dietary guidance for pregnant women remains unclear. This study aims to explore the capabilities of DeepSeek-R1 and ChatGPT-4.0 in generating dietary plans for pregnant women with different activity levels.

Personalized diet plans were generated using DeepSeek-R1 and ChatGPT-4.0. Through calorie calculation, Diet Quality Index-International (DQI-I) assessment, and cost analysis, the dietary quality and cost performance were evaluated.

The requested caloric targets in DeepSeek’s diet plans were superior to those of ChatGPT. All plans achieved a satisfactory DQI-I score (≥ 70). The “adequacy” score of DeepSeek-R1 was much higher (DeepSeek-R1 35.8 ± 0.7 vs. ChatGPT-4.0 33.9 ± 0.8, p < 0.001), while ChatGPT-4.0 performed better in the “moderation” aspect (ChatGPT-4.0 22.3 ± 2.2 vs. DeepSeek-R1 17.0 ± 3.4, p = 0.004). ChatGPT-4.0 demonstrated better performance in terms of cost-effectiveness (p = 0.017).

This study shows that DeepSeek-R1 and ChatGPT-4.0 can be helpful in providing personalized and reasonable dietary advice for pregnant women. In some aspects, such as food type adequacy, the emerging model “DeepSeek” performs better than ChatGPT.

Figures