Which OCR solution works best for your documents?
Comparing document-parsing systems on a benchmark of 2,937 pages across 16,288 tests, Opus 5.5 and GPT-6.1 Sol emerge as the top performers. Each system was run once on October 27, 2026, using the latest OpenRouter adapter and default reasoning. Opus 5.5 incurred 40 refusals, while GPT-6.1 Sol had 17, with no other failed pages in their respective runs.
Sonnet 5.5 had 7 refusals and 3 other failed pages, while GPT-5.6 Sol faced 17 refusals. Gemini exhibited the highest refusal count with 8, alongside 12 other failed pages. The scoring methodology involves evaluating the passage of each test across three families, resulting in an average headline score. No layout blocks are present in systems with no layout score, so no headline is assigned.
The repository includes tools to benchmark document-parsing systems using a unit-test scorer and dataset, as well as your own.
Written by urgent.news from Daily Dose of DS's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
