A Cairo-based artificial intelligence startup has released Horus 1.0-4B, a fully open-source large language model built in Egypt that outperforms several significantly larger global models on multilingual and Arabic-language benchmarks, in what developers describe as only the second open-source Egyptian AI model released in recent years.
TokenAI, founded by developer Assem Sabry, released the four billion parameter model this month under an MIT license. Despite Egypt graduating approximately 60,000 technology students per year and employing half a million workers in its ICT sector, homegrown AI models have been rare. The model is multilingual, optimized for Arabic language and cultural contexts, and available in seven variants to suit different hardware and deployment environments, ranging from a full 16-bit version requiring approximately 8 gigabytes to a compressed 2.3-gigabyte variant suitable for personal computers and edge devices.
On the MMLU benchmark, which tests multilingual knowledge across 57 academic subjects, Horus 1.0-4B scores 88%, outperforming Qwen 3.5-4B at 73%, Llama 3.1-8B at 69% and Gemma-2-9B at 71% — all models with substantially larger parameter counts.
On Arabic-specific benchmarks, Horus performs competitively. It scores 67% on ArabicBench, ahead of Qwen 3.5-4B at 65%, Llama 3.1-8B at 40% and Gemma-2-9B at 60%. On ERQA, an Arabic entity-rich question-answering benchmark, Horus scores 67% against Qwen’s 60%. Arabic mathematical reasoning remains a relative weakness: on AraMath, Horus scores 33% against Qwen’s 40%, an area the developer acknowledges for future improvement.
The model supports chain-of-thought reasoning, instruction following and tool use, and scores 84% on Terminal Bench, a command-line performance benchmark. It is available via Hugging Face and through TokenAI’s neuralnode Python framework.
A text-to-speech model called Replica is planned for imminent release, offering 20 voices across 10 languages including Arabic. TokenAI envisions Horus as the foundation of an Egyptian open-source AI infrastructure, with future models planned to expand capabilities while maintaining cultural alignment and Arabic language focus.
Horus arrives as Egypt builds AI momentum. In February 2026, the country’s IT industry development authority unveiled Karnak, Egypt’s first national large language model, positioned as the highest-ranking Arabic model in the 30 to 80 billion parameter range. Karnak powers several applications including an Arabic-language AI tutor, a legal assistant for citizens and small businesses, and healthcare AI tools for early disease detection. Abu Dhabi’s Mohamed bin Zayed University of Artificial Intelligence had previously released Nile-Chat, two open-source models built for the Egyptian Arabic dialect. With Horus, Karnak and Nile-Chat all available as open source, analysts say more Egyptian developers may now be inspired to build their own AI models.
