Skip to content

AI that runs where you are.

Razin builds AI that works on your phone, in your browser and on your own machines, in Arabic and English. Your data stays where it already is.

Apps, agents, models, and help using them.

Each product does one thing properly. Together they cover AI on the device, AI in the browser, and AI that reads documents.

Orb

App

Available now

Private, offline AI for Android. Chat, ask about photos and documents, and talk by voice, with no account and no cloud.

Models
Qwen3-VL 2B, Qwen3-VL 4B, MiniCPM-V 4.6, Gemma 4 E2B
Voice
On-device Whisper
Requirements
Android 10 or later; 6 GB RAM recommended
Price
One-time purchase. No ads, no subscription

Deputy

Agent tool

Developer preview

An agent in your browser that turns any page into a typed API for other agents. One call instead of screenshots and clicks.

Runs in
Your browser, in your logged-in session
Talks to
Any MCP client, such as Claude Code, Codex or Cursor
Source
On GitHub
Deputy demo

One form, two ways.

Deputy against Playwright MCP, filling the same seven-field booking form with the same prompt. Both succeeded.

Deputy against Playwright MCP on the same task
MeasureDeputyPlaywright MCPResult
Claude tokens133,366216,8881.63× fewer
Total cost$0.1861$0.27301.47× cheaper
Wall clock17.2 s26.1 s1.52× faster

Same seven-field form, same prompt, both through claude -p. Source: Deputy measurements

Razin OCR

Model

In development

General-purpose OCR for images and scanned PDFs, on your own hardware. We're making it especially good at Arabic, dots and diacritics included.

Input
PNG, JPEG and scanned PDF
Output
UTF-8 text, plus JSON with regions and reading order
Runs on
Your own CPU, with no network
Stage
Pretrained models; no fine-tuning yet
Quarterly report 2026
Order no. ABC-123
{ "text": "Quarterly report 2026", "reading_order": 0 }

Where pretrained models stand.

ConfigurationStrict CERMedian s/image
Tesseract fast, ara+eng52.80%0.199
Tesseract best, ara+eng42.03%0.288
PaddleOCR PP-OCRv5 + Arabic recognizer15.23%1.888

Arabic, 14 synthetic pages, CPU, 6 September 2026. Diagnostics, not accuracy on your documents.

Consulting

Service

We help teams run models privately, read Arabic documents, and connect agents to the web they already use.

  • On-device and private AI

    Choose, quantize and run models on phones, laptops and your own servers, with GPU and NPU acceleration and offline-first apps.

    Built on what we learned shipping Orb

  • Document AI, Arabic included

    OCR and extraction for documents in Arabic, English and mixed scripts, evaluated on your own pages before anything goes to production.

    Built on our work on Razin OCR

  • Agents and the web

    MCP servers, WebMCP, and turning the web apps you already run into tools agents can call safely.

    Built on what we learned building Deputy

  • Evaluation before investment

    Reproducible benchmarks on your data, so model, vendor and hardware choices rest on numbers.

    Built on our evaluation harness

Composed is the whole idea.

  • Runs where you are

    Models run on the phone, in the browser or on your servers. Your data stays put unless you decide otherwise.

  • Measured, not asserted

    Every number we publish comes with the method that produced it. If we haven't measured it, we don't claim it.

  • Arabic as a first language

    Arabic isn't a translation layer. We test on Arabic text, mixed scripts and right-to-left layouts from the start.

  • Say what isn't built

    Limits get listed as clearly as features. A developer preview is labelled as one.

Need AI on your own terms?

Tell us what you're trying to do. We'll tell you plainly whether we can help, and how we'd measure it.

hello@razin.ai
Talk to us