Nawah

Every AI model Newera runs is one we built, trained, and own. Arabic-native. Deployed in-Kingdom. Noforeign API in the path.
O W N E R S H I P

Renting a model means renting your capability.

No egress.
Inference happens where your data lives. Nothing is sent anywhere to be processed
No silent change.
Model versions are pinned to your deployment. Nothing shifts underneath a workflow that has already passed audit
No rental.
The weights run under your governance. Your capability does not depend on a commercial relationship staying favourable.
T H E  F A M I L Y

Four capabilities. One family. All ours.

The brain of the platform.

Every request routes through Conductor, which decides which model answers, what context it sees, andwhat it is permitted to do. It is where policy becomes enforceable rather than aspirational.
01 LLMs

Arabic and English language models for enterprise and government domains. 0
02 Speech / TTS  

Speech recognition and synthesis tuned for Saudi accent and dialect.
03 Computer Vision

Document, image, and video understanding, including Arabic document layouts
04 RAG

Retrieval over your own corpora with citation and access control enforced at document level.
L A N G U A G E
Arabic-native, not translated.
Generic models treat Arabic as a translation target. They are trained overwhelmingly on English, thenadapted. The result works for simple prose and fails on the things enterprises actually have — regulatorylanguage, dialect in customer conversation, mixed Arabic-English business documents, and domainterminology with no clean English equivalent.
Nawah is trained on Arabic as a first-class language.
G E N E R I C M O D E L S
English-first, Arabic adapted afterwards
Modern Standard Arabic only; dialect degrades
Latin-script assumptions in document layout parsing
Latin-script assumptions in document layout parsing
N A W A H
Arabic and English trained together
Saudi dialect handled natively
Terminology drawn from translated corpora
Arabic document layouts as a first-class case
E V A L U A T I O N
Test it against your own data.
Every model ships with an evaluation suite you run yourself, on your own infrastructure, against your owncorpora. We publish our benchmark results, but ours are not the ones that should decide your procurement

Run Nawah against your own corpus.