SoMark's 1.3B OXR tops olmOCR-Bench, OmniDocBench with pluggable layout module
pmttyji · reddit · 2026-11-07
The SoMark team released OXR-1.0, an open-source document parsing model on Hugging Face: just 1.3B parameters, pretrained from scratch with a recursive self-improvement (RSI) loop during post-training — a first in document parsing. It ranks #1 on olmOCR-Bench, OmniDocBench, and DocParsingBench, with a notably wide lead in table parsing.
- Recognizes 13 document element types (formulas, tables, chemical equations/structures, code blocks) with bounding-box coordinates; handles tough layouts like A3 exam papers, newspapers, vertical text, and handwriting.
- Ranks #1 in speed among open-source parsing models on RTX PRO 6000.
- First to make layout analysis independent and pluggable — swap the built-in layout model while keeping recognition and reading-order capabilities.
- Ships with an open-source ecosystem: SoMarkDown, SoPDF, and Python/JS SDKs.
More from Models
- 4B Nemotron 3 Nano claimed to beat frontier models on task-specific workflows — eliano · 2026-11-07
- Thinking Machines' Inkling 2 Small hits 81.7% on ARC-AGI-2, full weights coming next week — ziqiao_ma · 2026-11-07
- LithosAI claims fastest Kimi K3.1 Pro inference at 4.7 min per Intelligence Index task — JiaZhihao · 2026-11-07
- Applied Compute shares recipe for production RL with continual learning — ypatil125 · 2026-11-07
- Qwen3.8 27B writes 1,030-word replies, 6x longer than GPT-6 Sol's 174 — ArtificialAnlys · 2026-11-07
- China Telecom's Xing4.0-29B-A4B open-sourced: trained on Ascend NPUs, 75 on SWE-bench Verified — jacek2023 · 2026-11-07