Frontier AIData in Spain
OdiModel brings together the most powerful open-source models (Gemma, Qwen, GLM, Nemotron, DeepSeek) on dedicated NVIDIA HGX B200 infrastructure on Spanish soil.
Built for companies that can't afford data leaks
Analysis, documents and conversation in a single interface. Sovereign by design, not as an option.
Data in Spain, always
NVIDIA HGX B200 servers in certified datacenters on Spanish soil. No international transfers. No training on your conversations. No non-EU intermediaries.
Frontier open-source models
Always updated with the most powerful models on the market: Gemma, Qwen, Nemotron, GLM, DeepSeek. If you need a specific model that's not available, we deploy it within 24 hours. No vendor lock-in.
Text, vision and web search
Upload PDFs, images or Office documents. Live web search with source citation. A single interface for everything your team needs, no jumping between tools.
GDPR, ENS and ISO 27001 out of the box
Spanish company, European infrastructure. Data processing under European legislation. AES-256 encryption in transit and at rest. Per-account isolation at session, quota and storage levels.
Switch models, not vendors
The catalog is organized by use case. You pick what fits your task, not the other way around.
For chat
Everyday conversation, drafting, translation, summaries and quick lookups.
For reasoning
Deep analysis, math, logic, tasks that require chain-of-thought reasoning.
For vision
Vision-language: OCR, image analysis, extraction from scanned documents.
For coding
Code generation and debugging, agentic tasks, large-scale automation.
Need a specific model that isn't here? If it meets open-source conditions, we deploy it within 24 hours. The catalog is yours, not ours.
Built for sectors that don't tolerate compromises
If your organization operates under regulatory pressure (GDPR, ENS, professional secrecy), OdiModel is built for you.
Regulated sectors
Public administration, healthcare, defense, banking, insurance, legal. Organizations that can't send messages or documents to services outside the EU.
Regional administrations
Catalonia, the Basque Country, Galicia and other territories with co-official languages, plus public services and media operating in them.
Product and business teams
Non-technical users who need a reliable assistant for drafting, analysis, summaries, translation and search, without worrying about the model under the hood.
Technical and R&D teams
Developers, data scientists and ML teams who want to compare open-source models on the same prompt before integrating them into internal pipelines.
GPU Solutions customers
GPU Solutions ecosystem adding an assistant layer on top of the sovereign infrastructure they already consume, without closed-vendor licenses.
Media and broadcasting
Outlets and broadcasters publishing in multiple official languages, who need transcription, translation and summarization at broadcast quality on national soil.
Technical sovereignty, layer by layer
Every layer is selected so that data, compute and keys remain inside the EU end to end.
What security teams ask us most
OdiModel brings together in a single interface the most powerful open-source models on the market: Gemma-4-E4B, Qwen3-14B, Qwen3.5-9B, Qwen3.6-27B, Nemotron-VL-8B, Nemotron-30B, DeepSeek-V4-Flash and GLM-5.1-FP8 (754B). The user picks per conversation. The catalog grows continuously and, if you need a specific model that isn't available, we deploy it within 24 hours as long as it meets open-source conditions.
Yes, GDPR-native. All infrastructure is on Spanish soil, encryption keys are managed in the EU, there are no international transfers, and we don't train on customer conversations. Additionally, we hold ISO 27001 certification (EQA) and comply with the Spanish National Security Framework (ENS) at medium category.
In certified datacenters on Spanish soil. Specifically, on NVIDIA HGX B200 clusters with GPUs dedicated per enterprise customer, no sharing. Data, compute and key management remain inside the EU end to end.
No. Your conversations are not used to train models, are not stored for that purpose, and are not shared with third parties. They are processed securely and isolated on our infrastructure, accessible only to the account owner.
Only you. Conversations are archived in your account's side panel, accessible only to the owner unless legally required. Each customer operates on a cluster with isolated storage and quotas at account level. And you can delete any conversation or document at any time.
OdiModel is currently in public beta with a free, no-commitment plan. Features and models may change while we iterate with real users. Paid plans (individual, team and enterprise) will be announced when the beta closes. For enterprise customers with specific needs, we offer custom plans starting now.
Your company already uses AI.
Where is that data?
Free plan during the beta phase. Sign up in less than a minute. No credit card.
Free plan during the beta
No card. No commitment. Access to the full model catalog and HGX B200 infrastructure from minute one.
Want to know more?