Extend, a New York document AI company, has launched Operator-1, an autonomous agent for extracting data from complex documents. It combines document tools with persistent memory in a dedicated workspace, so it can extract data, check its work, and reuse methods that succeeded. Extend says Operator-1 set records on three benchmarks: 99.3% on LongArray-Extract, 99.88% on LongExtractionBench, and 96.57% on ExtractBench. Third parties published two of the three. The agent can write and run code, search across thousands of pages, and zoom into document images. It retries when a check fails. Operator-1 is available now in the Extend dashboard and API on all plans, at 5 credits per page. New users get 10,000 free credits.

Extend, a document processing platform for developers, has released performance results for its Light Parse engine on two open-source benchmarks. Light Parse achieved 90.5% field-level accuracy on RealDoc-Bench and 54.9% question-answering accuracy on Databricks OfficeQA Pro, whilst costing $0.00625 per page. The engine is designed for text-heavy, high-volume documents such as reports and contracts. Both benchmarks test whether AI systems can answer questions correctly using parsed document output. Extend reports Light Parse placed first among non-agentic parsers on both benchmarks. The company says customers in sectors including legal AI have halved their OCR spending after switching to its solution. Developers can select between Light Parse and the company's Performance Parse engine via API or Extend Studio.

株式会社ExtenD, a startup from the National Institute of Advanced Industrial Science and Technology, completed a seed funding round in June 2025. The funding, amount undisclosed, was led by Samsung Diamond Industrial Co., Ltd. and Saga Bank Capital & Consulting. The funds will be used for R&D and commercialization of diamond materials, talent acquisition, and business expansion. ExtenD aims to create new markets with its advanced diamond sensor technology and AI integration.