R&D Lab · From Granada with ♥
AI applications ready to deploy in your private environment
We build them and run them on our own platform in Spain. Try them today and deploy them in a dedicated environment for your organisation, with your data in Spain.
01
OdiModel
BetaThe AI assistant that keeps your data in Spain.
GPU Solutions Lab · Public beta 2026
The strongest open-source models (Gemma, Qwen, Nemotron, GLM) under one chat interface with native voice in Spain's official languages, on dedicated HGX B200 on Spanish soil.
02
Eridani
LivePublic research project running on our platform.
GPU Solutions Lab · Research · live
First Lab project released openly. Developed and operated end-to-end on GPU Solutions infrastructure in Madrid.
03
Odiverse
BetaEnterprise AI for finance.
Julio Sola · Founder
Chat with your fiscal, accounting, and treasury data. An AI assistant that understands your PnL, not one that promises dashboards.
04
Sustainable AI Benchmark
ResearchPublic energy-efficiency benchmark for AI models.
UGR + GPU Solutions · Research collaboration
Reproducible framework measuring the real cost in watts per token of a workload. Built with the Sustainable AI Infrastructure chair at the University of Granada.
We build and use these AI applications on our own platform in Madrid. You can try them today and deploy them in a dedicated environment for your organisation, with your brand, your users and your data in Spain.
- OdiModel: the best open models in a single chat, with voice in Spanish, Catalan, Basque and Galician (public beta).
- Odiverse: AI over your finance, tax and accounting data (beta).
- Eridani and the Sustainable AI Benchmark: research projects running on the same platform.
- Every application can be deployed on dedicated GPUs for your organisation, with data processed in Spain.
How it works
From demo to your own environment, in three steps.
- 01
Try it
A live demo or trial access, with no commitment, to check whether it fits the way your team works.
- 02
We deploy it for you
On dedicated GPUs in our Madrid cluster, with your brand, your users and your access rules.
- 03
We tune it
With your documents and your models, so it answers within your context, without your data leaving Spain.
Why
Ready-made applications on sovereign infrastructure.
Most organisations do not want to build an AI project from scratch: they want a tool their team can use next week. The problem is that nearly every off-the-shelf tool sends data to servers outside Europe, which rules it out for any area that handles sensitive information.
Our applications solve both at once. They are built and tested on the same platform we offer our customers, with NVIDIA B200 GPUs in a Tier III datacenter in Madrid, and they are deployed in a dedicated environment for your organisation. Data is processed and stored in Spain, never used to train models, and you can request zero retention.
And because the infrastructure is ours, the application is not a closed box. If tomorrow you need another model, more capacity or a connection to your systems, we do it on the same platform with the same team. If you would rather build your own application, our inference and dedicated GPUs are at your disposal.
Frequently asked questions
What is OdiModel?
OdiModel is an AI assistant that brings the best open models together in a single chat interface, with voice in Spanish, Catalan, Basque and Galician. It is in public beta, runs on our platform in Madrid and can be deployed for an entire organisation under its own brand.
Can I have OdiModel or Odiverse in an environment just for my organisation?
Yes. We deploy the application on dedicated GPUs in our Madrid cluster, isolated from other customers, with your brand, your users and your access rules. A standard dedicated environment is ready 72 hours after signature. If you need a different architecture, let's talk.
Which models do the applications use?
Open models served from our platform, including GLM 5.3, Qwen3.8, Gemma 4 26B, Nemotron 3.5 Lightning 30B and ALIA 40B, from the Barcelona Supercomputing Center, plus Whisper large-v3 for audio transcription. In a dedicated deployment you choose the models or use your own fine-tuned one.
Which languages do they support, including Spain's co-official languages?
OdiModel's voice works in Spanish, Catalan, Basque and Galician. The models are multilingual, so you can also write in English and other European languages, and they include ALIA 40B, from the Barcelona Supercomputing Center, developed with particular attention to Spanish and the co-official languages.
What happens to my data?
It is processed and stored in Spain, on our Madrid cluster. There are no subprocessors outside the EU, it is never used to train models and we sign a standard data processing agreement with optional zero retention. The platform is certified to ISO/IEC 27001:2023 and ENS (Spain's National Security Framework) Medium category.