Luma AI Opens Riyadh Office to Back Arabic-Native AI and HUMAIN Create
Feb 15, 2026 - Luma AI is setting up a dedicated office in Riyadh to accelerate HUMAIN Create, deepen regional partnerships, and push Arabic-native AI forward.
The office will act as a regional hub for client work, partnerships, and advanced model development. Luma plans to hire locally across engineering, go-to-market, and forward-deployed creative and technical roles.
Why this matters for builders
- Arabic-native models promise better tokenization, dialect coverage, and culturally accurate outputs compared to retrofitted multilingual models.
- Local presence shortens feedback loops with enterprises, regulators, and creators-useful for privacy, data residency, and brand safety needs.
- Expect faster iteration on evaluation sets that reflect GCC use cases: enterprise search, customer support, marketing localization, and media production.
HUMAIN Create and a Saudi-built foundation model
Luma AI will collaborate with HUMAIN on what's described as the world's first Saudi-built, Arabic-native foundation model, deployed through HUMAIN Create. The focus: culturally fluent outputs for creators and brands across the region.
Training and inference will be backed by Project Halo, a planned 2GW AI supercluster under development in Saudi Arabia. That scale points to serious throughput for pretraining, fine-tuning, and high-traffic inference once capacity comes online.
What to expect from the Riyadh hub
- Model work tuned for Arabic data quality: script normalization, diacritics handling, named entities, and mixed Arabic-English code switching.
- Benchmarks and evals that reflect local contexts and safety policies, not just global leaderboards.
- Enterprise integration paths: private fine-tuning, retrieval over first-party content, and deployment choices that respect data locality.
Hiring: signals for engineers and MLEs
- Distributed training and inference optimization (throughput, latency, and cost control at scale).
- Tokenizer and data pipeline engineering for Arabic corpora, plus alignment and red-teaming.
- Creative tooling: generative video and multimodal UX, asset pipelines, and review workflows for agencies and brands.
Agency workflows: Publicis Groupe Middle East
Under a separate agreement, Luma AI will be the preferred generative AI technology partner for Publicis Groupe Middle East. Publicis plans to integrate Luma's generative video and multimodal tools into creative and production workflows for faster, more localized campaigns.
- Practical stack moves: prompt libraries and templates, RAG over brand assets, content safety gates, and audit trails for approvals.
- For dev teams: make room for model/version governance, SDR/VTR thresholds for video output, and language QA for regional dialects.
How to plug in
- If you're building in MENA: line up Arabic datasets, evaluation sets, and red-team scenarios now so you can fine-tune on day one.
- Agencies: map your asset systems and compliance checks into a single orchestration layer to keep review cycles tight.
- Enterprises: decide early on deployment patterns (private cloud, VPC, or on-prem) to match data residency and latency targets.
Links and next steps
Upskilling a team on generative video, multimodal stacks, or evals? Browse the latest AI courses to get a focused shortlist.
Your membership also unlocks:
