MBZUAI's Jais 3 turns Arabic models into agents, from phones to GPU clusters

The six-model release pairs Arabic language models with speech recognition and publishes weights, training checkpoints and evaluation code under Apache 2.0.

        By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
        · Published 

Primary source: [Institute of Foundation Models on X](https://x.com/IFM_AI/status/2110641827778843010)

Why it matters #

Jais 3 combines Arabic speech, dialect-aware language models and tool use in an open release that can run from edge devices to data-center GPUs. Its performance claims are largely based on evaluations published by its developers, so adoption will turn on whether those capabilities hold up in real deployments.

The Institute of Foundation Models at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) released Jais 3 on October 15th, a six-model Arabic AI family built to handle spoken and written language and carry out tasks through software tools. The launch pairs models ranging from 0.9 billion to 375 billion parameters with Jais Speech, an Arabic speech-recognition system.…