Skip to content
All systems operational0 AI providers monitored, polled every 2 minutes
Live status

Institute of Foundation Models (MBZUAI)

ifm.mbzuai.ac.ae

The Institute of Foundation Models is the model-building arm of Mohamed bin Zayed University of Artificial Intelligence, launched on May 23, 2025 as a multi-site operation spanning Abu Dhabi, Paris, and a new Silicon Valley lab in Sunnyvale, California. It is the only lab in this directory whose primary target is a language other than English. On October 15, 2026 IFM and Inception42 published Jais 3, a fleet of six open models from 0.9B to 375B parameters under Apache 2.0, alongside Jais-ASR, an Arabic speech recognition model reporting a 22.85 percent word error rate across six public test sets under the Open Universal Arabic ASR Leaderboard protocol, which IFM says is ahead of ElevenLabs Scribe v2 and Speechmatics Melia 1 on that protocol. The flagship Jais 3 375B-A23B is a mixture-of-experts with 23 billion active parameters and a 512,000 token context window; the 32B, 36B-A4B, 7B, and 3.7B models share that window and the 0.9B model carries 128K. There is no hosted API and no token price: IFM ships weights, checkpoints, training recipes, logs, and a cleaned Arabic benchmark suite on Hugging Face and leaves serving to whoever wants it, which is a more reproducible disclosure posture than most of the labs above. IFM reports an Arabic average of 84.0 for the 375B model against 82.0 for Qwen3.5-397B, above Claude Sonnet 5 and behind GPT-5.6 Sol and Gemini 3.8 Flash, and says the 7B model reaches 79.4 on Arabic against 66.7 for the far larger Jais 2 70B. Infrastructure support came from Inception and Cerebras Systems.

Founded

2025 (Institute of Foundation Models)

Headquarters

Abu Dhabi, UAE

CEO

Richard Morton

Models

1 active

Key Products

Jais 3 375B-A23BJais 3 32BJais 3 36B-A4BJais 3 7BJais 3 3.7BJais 3 0.9BJais-ASR

Strengths

  • ✓Apache 2.0 weights across the whole Jais 3 fleet
  • ✓Built for Arabic first rather than evaluated on it afterwards
  • ✓Six sizes from 0.9B edge to 375B-A23B server scale
  • ✓Training recipes, logs, intermediate checkpoints, and a cleaned Arabic benchmark suite published
  • ✓512,000 token context on five of the six models

Institute of Foundation Models (MBZUAI) Models

ModelInput / 1MOutput / 1MContextCapabilities
Jais 3 375B-A23BFreeFree512Ktext, tool-use, code, reasoning

Prices per 1M tokens in USD. See the full pricing guide for detailed analysis.

Other Providers

Visit Institute of Foundation Models (MBZUAI)Compare ModelsAll Models