MX4 Atlas Documentation
Everything you need to build sovereign AI applications for the MENA region. Secure, compliant, and powerful.
Start Here
GuidesPopular Topics
ExploreFrequently Asked Questions
Do I need to change my application code to use MX4 Atlas?
No. MX4 Atlas is a drop-in replacement for OpenAI's API. Just change the base URL and API key—your existing code works immediately.
Is my data stored or logged by MX4?
Never. All data stays on your infrastructure. We don't retain logs, embeddings, or conversation history unless explicitly configured by your team.
What happens in an air-gapped deployment?
Atlas Runtime runs completely offline with zero internet connectivity. Model updates are distributed via USB or private networks. Activity journaling stays local.
Can I fine-tune models on proprietary data?
Yes. Fine-tuning runs on your infrastructure with your data remaining private. No data leaves your enclave during training.
What's the latency for inference?
Typical response time is 500ms–2s for chat depending on model size. For embeddings: <100ms. Check our benchmarks page for detailed specifications.
Need Help?
Docs by Role
For Developers
API guides, code examples, SDKs, and integration patterns.
For DevOps
Deployment, scaling, monitoring, and infrastructure setup.
For Security Teams
Data sovereignty, residency, activity journaling, and security practices.
For ML Engineers
Model fine-tuning, embeddings, benchmarks, and optimization.