Pricing
Every plan runs on EU bare-metal we own. Your documents never leave Europe. Your data never trains someone else’s model. Start free, scale to sovereign.
Upload your first documents. See what sovereign AI feels like. No credit card.
Unlimited documents. Full API access. The 72B model that actually reasons.
Dedicated database. Telemetry logging. The infrastructure your compliance team demands.
Your own fine-tuned model. Your GGUF weights. Infrastructure you could walk away with tomorrow.
All paid plans include a 7-day free trial. Annual billing saves 20%. Prices in Euro (EUR).
Enterprise Delivery
Our architecture team will audit your document landscape and deploy a production-ready sovereign engine for you. Fixed fee. No surprises.
Book The AI Dispatch →Pay As You Go
Every plan includes generous allowances. Go beyond them with transparent per-unit pricing — you only pay for what you use, and you’re never charged without consent.
from €0.05 / query
Beyond your plan allowance. Opt-in, never automatic.
from €0.15 / minute
NB-Whisper — Norwegian-optimized speech-to-text.
from €1 / GB / mo
Document and vector storage beyond plan limits.
from €250 / run
Train a dedicated model on your corpus. Corporate add-on, Sovereign included.
from €3 / source / mo
Automated web scraping for live data ingestion.
from €2 / 10K chunks
GPU embedding processing for large corpus uploads.
Add-ons are available on all paid plans. Usage is opt-in — you choose when to go beyond your allowance. Contact us for volume discounts.
Everything you need to choose the right plan for your team.
| Feature | Sandbox | Professional | Corporate | Sovereign |
|---|---|---|---|---|
| AI Queries | 100 / mo | 1,000 / mo | 10,000 / mo | Unlimited |
| Documents | 100 | Unlimited | Unlimited | Unlimited |
| Team Members | 1 | 3 | 50 | Unlimited |
| Storage | 500 MB | 5 GB | 50 GB | Unlimited |
| AI Model | Shared 27B | Shared 72B | Dedicated 72B | Own fine-tuned |
| REST API | — | |||
| Chat Widget | — | |||
| MCP Server Access | — | |||
| BNL Free Corpora | ||||
| Multiple Corpora | — | |||
| Web Scraper | — | 10 sources | 100 sources | Unlimited |
| Dedicated MariaDB | — | — | ||
| Telemetry Logging | — | — | ||
| Custom Branding | — | — | ||
| SSO / SAML | — | — | ||
| Corpus Fine-Tuning | — | — | Add-on | Included |
| GGUF Weight Ownership | — | — | — | |
| Dedicated Instance | — | — | — | |
| SLA Guarantee | — | — | 99.5% | 99.9% |
| Usage-Based Add-ons | — | Available | Available | Included |
| Support | Community | Dedicated | Dedicated + SLA |
Showing Professional plan features. Compare all plans on desktop.
The Sovereign Path
Most platforms lock you in deeper as you scale. We do the opposite. The further you go with CaveauAI, the more portable your investment becomes.
Upload 100 documents. Ask 100 questions. Every answer cites its source. Your data stays in Europe. No credit card, no commitment.
Unlimited documents. The 72B model that actually reasons. API access so your systems can talk to your corpus. This is where the flywheel starts spinning.
Dedicated MariaDB instance. Telemetry logging for every query. The compliance infrastructure your legal team requires before rolling AI into production workflows.
We fine-tune a model on your corpus and hand you the GGUF file. Run it on your hardware. Take it to another provider. Cancel and keep everything. This is what AI ownership actually looks like.
Your data never leaves Europe. Bare-metal servers in Germany & Finland.
Full data processing agreements. Privacy by design.
Your documents and queries are never shared across tenants.
Bare-metal infrastructure means 10x lower costs than cloud competitors.
Everything you need to know about CaveauAI pricing.
Yes. Upgrade instantly and your new limits take effect immediately. Downgrade at any time — changes apply at the next billing cycle. No lock-in contracts.
You'll receive a notification at 80% usage. When you reach your limit, you can either upgrade your plan or enable usage-based add-ons to keep going. If you do nothing, queries are paused until the next billing cycle. You are never charged without explicit consent.
Yes. All paid plans include a 7-day free trial with full access to every feature. No credit card required to start the trial.
All data is stored on EU/EEA bare-metal servers in Germany and Finland that we own. We never transfer data outside Europe. No cloud provider sits between you and your data. Full GDPR compliance with data processing agreements available.
We run open-weight models on our own hardware. Sandbox uses a shared 27B model for fast document search. Professional plans access the 72B model on shared GPU (NVIDIA RTX PRO 6000) for deeper reasoning. Corporate gets a dedicated 72B instance. Sovereign plans get a model fine-tuned on your corpus, with GGUF weights you own. No data is sent to OpenAI, Google, or any US-based AI provider — ever.
On the Sovereign plan, we fine-tune a model on your corpus and deliver you the GGUF weight file. You can run it on your own hardware, take it to another provider, or archive it. If you cancel tomorrow, your model walks out with you. No vendor lock-in.
Any industry with documents. Our first showcase is Norwegian family law, but the platform is industry-agnostic. Law firms, construction, healthcare, accounting, HR, compliance, education — if your business runs on documents, CaveauAI makes them searchable.
Every plan includes generous allowances for queries and storage. If you need more, you can enable usage-based add-ons at transparent per-unit rates (e.g., from €0.05 per extra query, from €0.15 per minute of audio transcription). Add-ons are always opt-in — you are never charged automatically.
Because gating documents is gating your value. The more of your business knowledge you upload, the smarter your AI gets. We gate compute instead — queries and model size — which is where the real infrastructure cost sits. Upload everything.
Yes. Professional, Corporate, and Sovereign plans support multiple corpora. Blue Note Logic corpora (Norwegian Family Law, EU AI Act & GDPR, EU CO2 Regulation) are free for all users. Future third-party marketplace corpora are priced individually.
Fine-tuning trains a dedicated AI model on your corpus data. Instead of just searching your documents at query time (RAG), the model learns your domain vocabulary, reasoning patterns, and terminology. The result is faster responses, lower per-query costs, and deeper domain understanding. Corporate plans can add fine-tuning, Sovereign plans include it with GGUF weight ownership.
Yes. Annual plans receive a 20% discount. Contact us for annual pricing on Corporate and Sovereign plans.
Every tenant is isolated. Your documents, embeddings, and queries are never mixed with other tenants. Corporate and Sovereign plans include dedicated MariaDB instances and telemetry logging for full audit trails.
Start in the Sandbox. Upload your first document in under 2 minutes. When you're ready, scale to sovereign.