Today for AI
HOT RADAR
llmHEAT 10.8°

AI CURATED HOT STORY · 10/27/2026

OXR Releases 1.3B Model, Tops Three Document Parsing Benchmarks

1 reports archived1 independent sourcesupdated 10/27/2026, 15:41:02
Synthesis & Latest Updates
1 Sources Cross-Validated

SoMark team open-sourced OXR, a lightweight 1.3B parameter model that ranks first on OmniDocBench, olmOCR-Bench, and DocParsingBench. It accurately parses 13 types of document elements including tables and formulas, outperforming PaddleOCR-VL-1.6 and MinerU2.5-Pro in speed while providing structured data for AI agents.

LATEST/SoMark team open-sourced OXR, a lightweight 1.3B parameter model that ranks first on OmniDocBench, olmOCR-Bench, and DocParsingBench. It accurately parses 13 types of document elements including tables and formulas, outperforming PaddleOCR-VL-1.6 and MinerU2.5-Pro in speed while providing structured data for AI agents.

Hourly heat curve

2 fully observed hours
Fewer than 3 fully observed hours; trend not plotted yet.

Coverage timeline

Total 1 reports · Latest first
  1. 量子位官网T2·78 pts

    OXR Releases 1.3B Model, Tops Three Document Parsing Benchmarks

    • OXR (1.3B params) ranks first on three major document parsing benchmarks, balancing accuracy and efficiency.
    • Supports recognition and structured output (Markdown/JSON) for 13 complex elements including tables, LaTeX formulas, and chemical structures.
    • Faster full-document parsing than PaddleOCR-VL-1.6 and MinerU2.5-Pro, suitable for building agent data pipelines.
OXR Releases 1.3B Model, Tops Three Document Parsing Benchmarks | Today for AI