Your AI models locally, on French GPUs
Run Llama 3, Mistral, Qwen or DeepSeek on dedicated GPU infrastructure with Ollama, installed and maintained by DINAO. A private LLM API, OpenAI-compatible — your prompts never leave France.
Was ist Ollama?
Ollama presents itself as « the easiest way to work with open models ». It allows you to run open-weight language models (Llama, Mistral, Qwen, Gemma and many others) directly locally, with installation and startup in just a few minutes.
Ollama exposes an OpenAI-compatible API, making it immediately usable by many existing tools (Open WebUI, n8n, Flowise…) and libraries. It handles parallel requests and integrates with development environments (Claude Code, etc.).
Its positioning emphasizes privacy: in local mode, work stays on your machine and works offline for critical use cases. A cloud option exists for scaling, but the core of the product remains the local execution of open models.
Hosten Sie Ollama bei DINAO
Ressourcenstufen kompatibel mit den Anforderungen von Ollama (min. 2 cœurs / 8 Go / 12 Go). In Frankreich gehostet, vollständig gemanagt.
- 4 dedizierte vCPUs
- 8 Go RAM
- 80 GB NVMe
- Tägliche Backups
- Gemanagt & überwacht von DINAO
- 8 dedizierte vCPUs
- 16 Go RAM
- 160 GB NVMe
- Tägliche Backups
- Gemanagt & überwacht von DINAO
2 verborgene Stufe(n) (unzureichende Ressourcen für diese App) : Découverte, Standard
Diese Anwendung nutzt KI
Der Container hostet die Anwendung, nicht die KI-Engine (die dedizierte GPUs erfordert) : Die Inferenz erfolgt extern, mit Ihrem eigenen Anbieter-Schlüssel. Bevorzugen Sie eine souveräne Engine — Mistral AI, NumSpot, Scaleway oder OVHcloud AI Endpoints (Frankreich, DSGVO) ; ein internationaler Anbieter (OpenAI, Anthropic…) nur, wenn eine spezifische Anforderung dies erfordert. Inferenz-Abonnements sind im Hosting nicht enthalten.
Technische Details
Sie fragen sich…
Which models can I run?
The entire Ollama catalog: Llama 3, Mistral, Qwen, DeepSeek, Phi, embedding models and your custom models. The maximum size depends on the VRAM of your tier.
Can I connect my existing applications?
Yes. Ollama exposes an OpenAI-compatible API: you simply change the base URL and key, and your tools (chatbots, RAG, scripts) work without rewriting.
Where is the data hosted?
On DINAO GPU infrastructure in France, in one of the available data centers. Your prompts and inferences do not leave the territory and are not used for any training.
Do I need technical skills?
To consume the API, some development knowledge helps. DINAO manages installation, the GPU, model downloads and updates. You can also use it via an interface like Open WebUI.
Can I change my plan or GPU?
Yes. You can upgrade at any time to run larger models, and your custom models remain yours — no lock-in.