YAML vs Markdown vs JSON vs TOON: Which Format Is Most Efficient for the Claude API

Summary of Findings

The study evaluated the impact of input format on the performance and cost efficiency of Anthropic's Haiku language model across various scenarios. Key findings are summarized below:

Accuracy by Scenario:

• Products (Catalogs): Markdown (TXT) performed best with 70.2% accuracy.
• Tasks (Roadmaps): JSON was most accurate at 71.0%.
• Hierarchy: YAML showed the highest accuracy of 92.9%, while Markdown lagged significantly at 57.1%.
• API Docs: Both JSON and YAML achieved top accuracy of 85.7%.

Accuracy by Size:

• Moving from small (S) to medium (M) data volume resulted in a significant drop in accuracy, highlighting the importance of reducing input size before optimizing format.

Cost Efficiency:

• JSON was the most expensive format, with an average cost per request of $0.0026.
• TOON was the least costly at $0.0010 per request.
• Transitioning from JSON to TOON resulted in a 62% reduction in costs, saving approximately $162 per month for

Read the full article at DEV Community

---

Want to create content about this topic? Use Nemati AI tools to generate articles, social posts, and more.