PYXON AI is PYXON's research division — a team of scientists and engineers building AI that outperforms big tech on Arabic-language challenges.
R&D labs that benchmark, build, and ship AI systems —not slides— starting with Arabic and specialized workloads. Reasoning search, on-device RAG, AI detection, poetry benchmarks: proven in research, ready to try.
PYXON AI Labs is the research division of PYXON where scientists and engineers design real-world intelligence systems.
Focused on complex challenges such as Arabic language understanding, the Labs builds AI that performs where conventional systems fall short.
This is where intelligence is not just applied — it is created, tested, and refined.
The first Arabic search engine that actually understands what you mean — not just what you typed. Independently benchmarked to outperform OpenAI, Google, and Cohere on Arabic semantic retrieval.
Read the article →Scroll to see demo
Why pay for a 7B-parameter cloud model when a 1.5B model on your phone can outperform it? Our on-device RAG runs fully offline, beats big-tech baselines, and keeps your data on your device.
Read the article →Scroll to see demo
Map 120+ GPUs from consumer cards to hyperscale AI infrastructure — paste your terminal output, match your hardware, and see which LLM sizes fit for inference and QLoRA fine-tuning.
Scroll to see tier map
A ~1.6 GB Arabic instruction model for edge and on-device AI — QLoRA fine-tuned on Qwen2.5-1.5B, runnable offline with Ollama. Pairs with on-device RAG when accuracy must match larger cloud systems.
Scroll to see model specs
98.4% accuracy detecting AI-generated Arabic text — outperforming GPTZero and OpenAI's own classifier. Built natively for Arabic from day one. Not translated. Not adapted. Engineered.
Scroll to see demo
PYXON AI is powered by a growing community of AI scientists, university students, software engineers, and industry specialists from across the MENA region and beyond. Our labs are where ideas are tested, benchmarks are challenged, and production-grade models are shipped. We welcome external contributors.
A culture of curiosity, ownership, and impact — across MEA.
Pyxon Governance scores LLMs and RAG systems against country AI rules (e.g. Saudi Arabia / SDAIA, UAE, EU) using config checks, live safety probes, and policy evidence—so teams can prove compliance before deployment. Available as pip install pyxon-governance and npm install @pyxon/governance.
75 randomized AI governance questions. After each answer you see the correct choice and explanation — then submit your contact when you finish.
If you are not yet in AI and want to join Pyxon, this interactive roadmap guides software engineers from foundation to enterprise AI delivery.
Scroll to view interactive roadmap
Whether you're looking for a role inside PYXON or to contribute as an external scientist or student — there's a door for you.
Talent & hiring: hr@pyxon.com
View open roles →We're looking for scientists and contributors to join PYXON Labs.
Or email info@pyxon.com