Central Brain™ system requirements and benchmarks

What is confirmed, and what should you check?

The product team confirmed Windows 10 or later with a 2 GB RAM minimum, and Apple M1 or newer for Mac, on October 17, 2026. The 2 GB figure is a Windows minimum, not a promise that large document collections or OCR fit comfortably in that memory. An exact minimum macOS version and Mac RAM requirement were not supplied. Check the installer and release instructions for your computer.

Plan storage for originals, prepared output, project state, memory and the local vector index, plus your backups. Keeping these components visible gives you a practical way to monitor project growth. Test with representative files, then use the observed output size and available disk space to plan your next project.

What runs locally, and what needs a connection?

Central Brain™ prepares supported PDF, DOCX, TXT, Markdown, CSV and XLSX files and searches them with local embeddings by default. Images, audio and video have metadata-only indexing. OCR availability depends on the build; confirm that feature before evaluating scans. Results carry source and location details where available.

Preparing, indexing, searching and memory work without a network connection. Keygen activation and a license check every 14 days need connectivity; Team and Enterprise can obtain a build without an online license check. A cloud-hosted assistant and any enabled cloud service have their own connection needs. Connected AI tools receive the project context they retrieve.

The product team reports over 10,000 PDF pages chunked in under 15 minutes

On October 17, 2026, NeuroAIgent reported chunking more than 10,000 PDF pages in less than 15 minutes. This vendor-reported, chunking-only result demonstrates the scale of a preparation run observed by the product team. Hardware, corpus, OCR settings and repeated-run measurements were not supplied, and the result has not been independently verified.

Measure chunking, OCR if used, embedding, indexing and retrieval separately to understand your complete workflow. Central Brain™ brings those preparation and retrieval tasks together with persistent project memory, so a useful evaluation follows the files all the way from a chosen folder to source-backed context in your assistant.

A reproducible Central Brain™ benchmark you can run

Use a documented corpus and known-answer questions. Measure local preparation, local retrieval, memory recall and assistant responses as separate operations.

1. Record the Central Brain™ build, OS, processor, RAM, free disk space, storage type, embedding configuration and OCR setting. Record the connected client and version separately.
2. Choose an authorized test corpus. Record file counts by format, total bytes, PDF pages, scanned pages and workbook complexity. Keep an unchanged copy so later runs use the same inputs.
3. Time initial preparation and indexing from start to completion. Record failed files and index/output sizes; speed without usable coverage is an incomplete result.
4. Create 20 questions with expected source files and locations. Include exact phrases, paraphrases, a question absent from the corpus, a spreadsheet question and a scan if OCR matters. Keep the same questions for every run.
5. Run each query three times with the same search settings. Record each local search time, whether the expected source appears, its position and whether you can inspect its location. Report the median and range, identifying first-run and warm-run conditions.
6. Change one document, add one and remove one, then measure the refresh and inspect the result. Save a non-sensitive decision, restart the client and test memory recall in a fresh session.

How to interpret the results

For each known-answer query, record whether the expected source was found and whether its surrounding text supports the answer. A returned snippet alone does not prove accuracy. Review spreadsheet values against the workbook and OCR text against the scanned page. Record failures rather than silently excluding difficult files.

Keep your measurement sheet with its build and corpus details. Compare runs with the same stages and conditions, and assess retrieval quality alongside throughput. A repeatable project gives you a useful basis for choosing a plan and sizing a larger workflow.