Unstructured alternatives

+ Produces natural, editable prose and follows nuanced instructions over very long documents better than most rivals.
− Pro usage limits can run out mid-session on long conversations, and it has no built-in image generation.

+ Reads a table as a table, and does it on your own machine with nothing sent anywhere.
− Heavy scanned documents want a GPU; CPU-only runs are slow on long PDFs.

+ Beats the old OCR engines on messy scans and non-Latin scripts, locally.
− The licence is not plain open source above a revenue line - read it before shipping.

+ Answers are grounded strictly in your uploaded PDFs, Docs, URLs and YouTube videos with citations, and Audio Overviews turn them into a two-host podcast.
− Sources are capped per notebook and the free tier limits daily queries and audio generations.

+ Holds up on financial statements and forms where plain text extraction turns to soup.
− Per-page pricing and a cloud round trip; large archives add up fast.

+ Very fast responses and EU-hosted data handling, with a cheaper Pro tier than the US incumbents.
− Smaller model ecosystem and fewer integrations, and answers lag the top US models on hard reasoning tasks.

+ Still the most accurate thing there is at equations, which general OCR gets wrong.
− Narrow by design - a plain business PDF is better served elsewhere.

+ Available almost everywhere, including an API for developers who want document Q&A inside their own product.
− Answer quality is middling on long documents and the free tier is restrictive on uploads and questions.

+ The Copilot sidebar explains equations, tables and jargon in a paper you are reading without leaving the PDF.
− The free tier is limited and the interface pushes upgrades aggressively.

+ Zero setup: drop a PDF and start asking, with page citations for every answer.
− Free plan caps file size and page count, and long or scanned PDFs often produce vague answers.