Paris-based AI startup Mistral has introduced a new language model, Mistral Saba, which focuses on Arabic-speaking countries. The model, launched on Monday, is tailored to handle Arabic interactions more effectively, setting it apart from Mistral’s general-purpose models.

With 24 billion parameters, Mistral Saba is compact compared to other large language models (LLMs). While smaller models often perform faster, larger ones are generally considered more intelligent. However, according to Mistral’s internal tests, Saba outperforms its similar-sized counterpart, Mistral Small 3, when dealing with Arabic content.

Interestingly, the model also shows strong capabilities with some Indian-origin languages, particularly Tamil and Malayalam, thanks to cultural and linguistic overlaps between the Middle East and South Asia.

Mistral’s latest release reflects its strategic focus on the Middle East, with plans to use Saba to attract customers in the region. The company envisions Saba being used for tasks like conversational AI and content generation in Arabic, while also serving as a foundation for fine-tuned models tailored to specific industries.

Mistral Saba can be accessed through the startup’s API or deployed on-premise—a feature designed to appeal to industries like finance, healthcare, and energy that prioritize data privacy.

Mistral has garnered significant funding from U.S.-based investors such as Lightspeed Venture Partners, Andreessen Horowitz, and Salesforce Ventures. However, with shifting geopolitical dynamics, the company is now eyeing potential investment from the Middle East. The release of Saba could play a key role in attracting regional backers, helping Mistral position itself as a global alternative to American and Chinese AI firms.

The launch of Mistral Saba also highlights the company’s commitment to multilingual AI. Mistral, known for its open-weight Mistral 7B model, has emphasized its dedication to supporting diverse languages. The company hinted that future models will focus on other regional languages as part of its ongoing expansion.