Tarjama’s Pronoia V2 Surpasses Global AI Giants in Arabic LLM Performance 

Tarjama, the UAE-based leader in Arabic language technology, has unveiled Pronoia V2, its latest Large Language Model (LLM) designed to revolutionize Arabic and multilingual enterprise applications. With this launch, Tarjama aims to address long-standing gaps in Arabic AI performance while setting new global benchmarks in linguistic accuracy, operational efficiency, and enterprise-readiness.

Backed by over 16 years of linguistic and technological expertise, Pronoia V2 significantly outperforms globally recognized models on Arabic benchmarks, including OpenAI’s GPT-4o and leading open-source alternatives. The new model represents a major leap in Arabic natural language processing and further strengthens the region’s position in the global AI race.

A Leap in Performance and Efficiency

Pronoia V2 scored an average of 76.8% across Arabic evaluation benchmarks, substantially ahead of:

• GPT-4o (58.05%)

• DeepSeek 70B (53.6%)

• Cohere 7B (55.3%)

The model also posted exceptional results on task-specific benchmarks, including 92.4% on MT ARC Easy and 83.4% on MT Openbook QA, demonstrating its strong contextual understanding and precision in specialized fields such as legal, medical, and business domains.

"Arabic has long been underserved by AI technologies," said Nour Al Hassan, Founder and CEO of Tarjama. "With Pronoia, we’re changing that narrative—delivering a solution that is not only more accurate and secure, but also faster and far more cost-efficient than anything else on the market."

Engineered for Business

Unlike many large-scale models that require substantial infrastructure, Pronoia is built to run on just one GPU, offering a uniquely efficient architecture without compromising performance. This efficiency, combined with advanced enterprise features, makes the model especially attractive for organizations operating in Arabic-speaking markets.

Key enterprise capabilities include:

• End-to-end data encryption

• Flexible cloud and on-premise deployment

• Custom model training options

• Support for Arabic and English text, audio, and structured data processing

Pronoia is already enabling enterprises to automate key functions such as contract review, multilingual customer service, regional content creation, and real-time document processing.

"We’ve seen firsthand how businesses struggle with managing Arabic content at scale," Al Hassan added. "Pronoia transforms that challenge into a competitive advantage—turning tasks that once took hours or days into processes completed in minutes."

A Strategic Vision for Arabic AI

Founded in 2008, Tarjama has long focused on elevating Arabic language technology. With Pronoia V2, the company reinforces its commitment to building AI solutions that reflect the linguistic and cultural nuances of the region.

According to Iyad Ahmad, Tarjama’s Chief Technology Officer, the success of Pronoia V2 challenges the conventional wisdom that model size is the primary determinant of performance.
"We’ve proven that a well-architected, precisely trained model can outperform much larger systems. Size doesn’t always equate to strength—it’s about how intelligently you build."

Looking Ahead: Pronoia Vv32

Tarjama is currently developing Pronoia Vv32, the next iteration of the model, with ambitions to push performance even further. The upcoming version targets an 80%+ average across Arabic evaluation benchmarks and a COMET score above 94, signaling best-in-class machine translation quality.

As demand for AI-driven solutions continues to rise in the MENA region, Tarjama’s innovations are poised to play a pivotal role in shaping the future of enterprise AI. With Pronoia, the company is not only addressing a longstanding technical gap—it’s also redefining the standards for what Arabic AI can achieve.