On October 15, 2026, the Institute of Foundation Models and Inception42 introduced Jais 3, a new fleet of Arabic-native AI models and speech recognition tools. Built on the K2 Horizon foundation, the release includes six open models and Jais Speech.
Model Fleet and Capabilities
The Institute of Foundation Models (part of Mohamed bin Zayed University of Artificial Intelligence) and Inception42 introduced Jais 3, a fleet of models built for Arabic. The new fleet pairs six open language models, ranging from 900 million to 375 billion parameters, with Jais Speech, a speech recognition model designed for Arabic and its regional dialects.
According to the creators, the models understand Modern Standard Arabic and regional dialects, cultural context, and code-switching between Arabic and languages like English or French. Jais 3 can perform practical tasks, follow complex instructions, and support multi-step agentic workflows.
Availability and Deployment
Jais 3 and Jais Speech are available today on Hugging Face, with deployment support for vLLM and SGLang. The models share a context window of 512,000 tokens and were created in collaboration with Cerebras Systems.
Alongside the models, IFM is releasing audited Arabic benchmarks, evaluation code, and training checkpoints. Enterprises and governments can adapt the models to meet sovereign AI requirements.
Key facts and where they come from
- IFM and Inception42 introduced Jais 3 on October 15, 2026.
The Institute of Foundation Models (part of Mohamed bin Zayed University of Artificial Intelligence) and Inception42 today introduced Jais 3, a fleet of models built for Arabic.
- The fleet includes six open language models ranging from 900 million to 375 billion parameters.
The new fleet pairs six open language models, from 900 million to 375 billion parameters, with Jais Speech, a speech recognition model for Arabic and its dialects.
- Jais 3 and Jais Speech are available on Hugging Face with vLLM and SGLang deployment support.
Jais 3 and Jais Speech are available today on Hugging Face, with deployment support for vLLM and SGLang.
Read the original from Company press releases (PR Newswire) →
