Source: Unite.AI
You.com launched the Extract API on November 17, 2026, giving developers one endpoint to fetch a web page or PDF and return its contents as Markdown or HTML. The company reported that its default PDF mode scored at least 80% higher than Exa Contents and Parallel Extract in a five-dimension evaluation.
Extract is built for agents that need an entire filing, contract or research paper after a search rather than a snippet, although it can also be called directly when the URL is already known. Requests use an API key, and responses identify whether the material came from a cache or a live fetch.
Retrieval and Parsing Controls
The API reference provides three retrieval settings. The default, blend, uses cached content when available and otherwise crawls the page live; fetch always crawls live, while cache returns stored content only and omits the content field when no copy exists. Requests have a 30-second end-to-end timeout by default, configurable from 2 to 100 seconds, after which the endpoint returns a 504 error.
PDF processing is controlled separately by parse_effort. Standard, the default, sends complex pages such as charts and images through an advanced model; lite extracts only the content embedded in each page. You.com says the first approach is intended for tables, charts, forms and scanned pages, while lite suits digitally created, text-based documents such as contracts. Neither setting changes web-page output or pricing.
Benchmark Methodology and Results
To test PDF quality, You.com ran both Extract modes, Exa Contents and Parallel Extract over 3,782 pages in 12 ParseBench-2 sets. Each service received the same URL, ParseBench’s grader scored the outputs, and any document a service failed to return received zero. You.com reported its largest advantages in tables and charts: Extract standard scored 74.9 and 76.6 in those categories, respectively, against 22.8 and 0.9 for Exa and 0.1 and zero for Parallel.
Speed moved in the opposite direction: You.com reported median latency of 11.0 seconds for standard and 2.6 seconds for lite, compared with 1.3 seconds for Exa and 1.8 seconds for Parallel. Parallel’s minimum cache age of 600 seconds means some of its responses may have been cached.
The parsing scores also cover a narrower evaluation than the full benchmark. You.com excluded the coordinate-based layout evaluation because Extract returns parsed content rather than page-element coordinates. LlamaIndex’s public leaderboard covers 56 methods, 4,317 pages and 13 datasets across six dimensions, with the overall score calculated as the plain mean of those dimensions. The headline comparison therefore describes You.com’s scoped test, not the full ParseBench-2 overall ranking.
Pricing and Contents API Transition
Lite costs a flat $0.90 per 1,000 URLs. Standard uses that URL rate and adds $0.50 per 1,000 pages only when the fetched file is identified as a PDF.
Based on its reported 26-page median across the billions of PDFs it has processed, You.com puts the cost of 1,000 typical documents at $13.90 on standard, against the flat lite rate, and lists Exa and Parallel at $1.00 per 1,000 URLs regardless of length. The deeper standard mode therefore carries a substantial multi-page premium, while lite undercuts those rivals for text-based PDFs.
Requests that fail on You.com’s side are not billed, and error responses distinguish blocked URLs, failed requests and pages without content so pipelines can decide whether to retry. Zero data retention is available for sensitive documents at no additional charge, but customers must contact You.com to enable it.
Developers can use Extract on web pages through You.com’s official MCP server for free, without a key or authentication. Direct API sign-ups include $10 in credit, or $100 with a work email; You.com estimates the larger amount covers about 7,000 typical PDFs on standard. The older Contents API remains supported for existing users but is no longer in active development, with deeper PDF parsing and parse_effort reserved for Extract.
The dividing line for retrieval pipelines is the document itself. If the answer sits in embedded text, lite preserves a flat per-URL cost and faster response; if it sits in a chart, table or scanned page, You.com’s reported parsing gains come with higher latency and, for long PDFs, a much higher bill.

