LlamaIndex releases OCRia-1.5, setting state of the art on ParseBench-2 for tables and charts

·2026.11.18 09:00

Key point

The new 4B and 2B models improve parsing accuracy for complex tables, charts, and forms while enhancing visual grounding precision.

1 / 5

Details

  • LlamaIndex introduces OCRia-1.5, a document parsing model upgrade that sets the state of the art on ParseBench-2 for tables, charts, and formatting.
  • Table scores on ParseBench-2 rose from 86.5 to 90.1 in the Agentic tier and from 83.2 to 87.5 in the Cost Effective tier, with larger gains on tables featuring spanning headers.
  • Visual grounding improved significantly, with the Cost Effective tier score rising from 77.0 to 88.5 and the Agentic tier increasing by 4.1 points.
  • OCRia-1.5-4B runs the Agentic tier at 10 credits per page, while the new OCRia-1.5-2B runs the Cost Effective tier at 3 credits per page for high-volume parsing.

Model Tiers and Pricing

  • OCRia-1.5-4B is the upgraded flagship model powering the Agentic tier.
  • OCRia-1.5-2B is a smaller model with half the parameters of OCRia-1, powering the Cost Effective tier.
  • Both models are live in LlamaParse immediately upon release.

Performance Improvements

  • Chart scores increased from 74.9 to 76.8 in Agentic and from 67.3 to 73.0 in Cost Effective.
  • Forms scores rose by 2.6 points in Agentic and 4.2 points in Cost Effective, ensuring values stay attached to correct fields.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.