{"@context":"https://schema.org","@graph":[{"@type":"Service","name":"Yorizon Factory AI","description":"Managed AI platform with OpenAI-compatible inference, AI Workbench, and open-weights hosting on B200 - European, without Cloud Act.","provider":{"@type":"Organization","name":"Yorizon","url":"https://yorizon.com"},"areaServed":{"@type":"Place","name":"Europe"},"serviceType":"AI Inference Platform","category":"AI Platform","url":"https://yorizon.com/products/yorizon-factory-ai"},{"@type":"FAQPage","mainEntity":[{"@type":"Question","name":"Which models are available?","acceptedAnswer":{"@type":"Answer","text":"Phase 1: Llama 3.1 70B, Mistral NeMo, DeepSeek-R1. Phase 2: Llama 3.1 405B and Flux.1 (Image). Open-weights models with hosting license, European models preferred."}},{"@type":"Question","name":"Is the API OpenAI-compatible?","acceptedAnswer":{"@type":"Answer","text":"Yes, fully OpenAI-compatible REST API for chat, completion, and embedding requests. Existing applications can be integrated seamlessly."}},{"@type":"Question","name":"Where does my data stay?","acceptedAnswer":{"@type":"Answer","text":"Data never leaves the Yorizon data center - no US Cloud Act. Addresses GDPR Art. 9, NIS2, and the EU AI Act."}},{"@type":"Question","name":"What does usage cost?","acceptedAnswer":{"@type":"Answer","text":"Pay-per-use per million tokens or requests. Token latencies, throughput per model, and guaranteed quotas are agreed with each customer individually."}},{"@type":"Question","name":"Can I host my own models?","acceptedAnswer":{"@type":"Answer","text":"Hosting of fine-tuned open-source models is optionally available (specification as per offer). AI Workbench with PyTorch, JAX, and vLLM for fine-tuning is available."}}]}]}
Yorizon Factory AI
Yorizon’s managed AI platform - Inference API, AI Workbench and model hosting on European GPU infrastructure.
Software developers, data teams, and business units that want to run open-source models and RAG applications in production—without their own GPU infrastructure, without model hosting complexity, and without data leakage to US cloud providers.
What Yorizon delivers
Inference endpoints for Llama 3.1, Mistral, DeepSeek
OpenAI-compatible REST API
AI Workbench with PyTorch, JAX, vLLM
Pay-per-use per million tokens
Hosting of your own open-source models by arrangement
Multi-tenant and white-label mode
Architecture & Technology
NVIDIA NIMs on B200 nodes under Direct Liquid Cooling. Multi-node inference via RoCEv2. OSISM (Sovereign Cloud Stack) with Ceph and Neutron.
Security & Sovereignty
Data never leave the Yorizon data center
EU-only data residency, no US Cloud Act
Addresses GDPR Art. 9, NIS2, EU AI Act
Tenant separation at the identity, network, and model levels
Audit logs for API calls
Service Level
99.9% monthly availability
Token latency and throughput per model on request
Tiered service credits
Which models are available?
Phase 1: Llama 3.1 70B, Mistral NeMo, DeepSeek-R1. Phase 2: Llama 3.1 405B and Flux.1 (Image). Open-weights models with a hosting license, European models preferred.
Is the API OpenAI-compatible?
Yes, a fully OpenAI-compatible REST API for chat, completion, and embedding requests. Existing applications can be seamlessly integrated.
Where is my data?
Data never leave the Yorizon data center - no US Cloud Act. Addresses GDPR Article 9, NIS2 and the EU AI Act.
How much does it cost to use it?
Pay-per-use per million tokens or requests. Token latencies, throughput per model, and guaranteed quotas are agreed individually with the customer.
Can I host my own models?
Hosting finely tuned open-source models is optionally available (details on request). An AI workbench with PyTorch, JAX, and vLLM for fine-tuning is available.
