Read every paper
as if it were written
in your language.
Paste an arXiv link. Get back a bilingual PDF and a side-by-side reader with a chat that has actually read the paper. No marketing tour. Just the article.
We introduce SWE-Bench Pro, a substantially more challenging benchmark that builds upon the best practices of SWE-Bench but is specifically designed to capture realistic, complex, enterprise-level problems beyond its scope.
Мы представляем SWE-Bench Pro — значительно более сложный бенчмарк, который строится на лучших практиках SWE-Bench, но специально разработан для решения реалистичных, сложных задач корпоративного уровня.
We didn't build a chatbot that summarises papers. We built a reader that respects them — formulas stay LaTeX, figures stay figures, every citation keeps its number.
From a link to a finished read.
Paste the link
Drop in any arXiv URL — abs, pdf, or e-print. We pull metadata, count pages, and tell you the cost before anything starts.
We translate the article, not just the words
Math stays LaTeX. Figures stay figures. Citations keep their numbers. Each paragraph is matched 1:1 between languages so the columns scroll in lockstep.
Read with a chat that knows the paper
Highlight any passage to ask about it. Answers cite the section they came from, so you can verify everything you read.
Pairs that mostly don't exist anywhere else.
Tuned per-pair for technical writing — math, code identifiers, citations, and domain idiom.
Free for the first three. Then $15.
Free covers casual reading. Pro is fair-use unlimited, every language, AI chat included. Premium adds the heavy model and DOCX/EPUB export.