Free Forever for Developers. Enterprise Support When Needed.
No cloud tokens. No per-seat billing. Total ownership of your models, data, and infrastructure.
Complete access to the local model adaptation foundry. Unlimited models, unlimited runs, 100% private.
- Epistemic Data Compiler with MI Guard
- Pre-flight VRAM hardware predictor
- 5-step self-healing OOM auto-recovery
- Multi-layered evaluation & drift bounds
- Mission Control real-time web UI
- 1-click GGUF export & Ollama deployment
- Community Discord & GitHub support
For security, compliance, and engineering organizations requiring air-gapped deployments and custom hardware acceleration.
- Everything in Community Edition
- Air-gapped on-premise installation support
- Custom enterprise data connectors (SAP, Epic, DB2)
- Multi-GPU & Slurm / Kubernetes cluster scaling
- Dedicated ML engineering advisory & SLA
- Regulatory audit & SBOM certification assistance
Frequently Asked Questions
Everything you need to know about licensing, data privacy, and hardware requirements.
Is MoroAI really 100% free and open source?
Yes. MoroAI is fully open-source under the permissive Apache 2.0 license. You can use it commercially, adapt models on your private servers, modify the code, and deploy without paying any license fees.
Does my training data or telemetry ever get sent to cloud servers?
Never. MoroAI is strictly local-first. All compilation, training loops, checkpoints, loss telemetry, and Ollama deployments occur entirely on your local machine or private VPC.
What hardware do I need to run MoroAI?
MoroAI is optimized for single consumer GPUs. You can train 1.5B parameter models on an 8GB VRAM GPU (e.g. RTX 3060, 4060, or Apple Silicon M-series) in under 30 minutes. Larger 7B–14B models run smoothly on 16GB–24GB VRAM cards.
What does Enterprise Support include?
Enterprise Support is designed for regulated organizations (financial institutions, defense contractors, healthcare) that need custom data compiler connectors, air-gapped on-premise deployments, SLA-backed response times, and multi-node cluster optimization.