€1 an hour. Prepaid. No subscription.
Pay only for the minutes a machine is actually running — €1/hour for the Max tier (frontier models), €0.50/hour for the Core tier (everyday models), from prepaid credit. Unlimited colleagues run their own machine at once. Or buy a box outright for €6,000.
Pryvan Core
Everyday models · paid by the hour
A private AI machine for everyday work — drafting, summaries, document Q&A and coding help.
- Compute rate
- €0.50 = 60 minutes
- Good for
- Drafting, summaries, doc Q&A, coding
- Every session
- A fresh, anonymous machine
- When you finish
- Wiped — nothing is kept
- Fast everyday open models for chat, writing and coding
- Private chat + document Q&A with citations
- A brand-new machine every time you connect
- All session data deleted when you disconnect
- Email support
Billing: Billed by the minute — only while the machine is on
Join the waitlistPryvan Max
PopularFrontier open models · paid by the hour
The powerful tier for coding, agents, long documents, vision and deep reasoning.
- Compute rate
- €1 = 60 minutes
- Good for
- Coding, agents, long docs, vision
- Every session
- A fresh, anonymous machine
- When you finish
- Wiped — nothing is kept
- Runs the strongest open models available today
- Long documents, vision, agentic and reasoning workflows
- A brand-new machine every time you connect
- Everything in Core
- Priority support
Billing: Billed by the minute — only while the machine is on
Join the waitlistOn-prem
Your office · you own the hardware
Prefer to own it? Buy a comparable machine outright and run it in your own office.
- Cost
- €6,000 + VAT, one-time
- People at once
- 1 at a time
- Upkeep
- Yours to house & maintain
- Lifespan
- Not top-of-the-line after ~3 years
- A capable AI server shipped to your office
- One person uses it at a time
- You handle setup, cooling, updates & troubleshooting
- Runs (and draws power) whether used or not
- Not top-of-the-line after ~3 years, no upgrades included
Billing: You own it: electricity, maintenance and troubleshooting are on you
Talk to usYou pay only for the minutes a machine is actually running — €1 per hour on Max, €0.50 per hour on Core, drawn from prepaid credit. No per-token metering, no monthly minimum. Pause the machine and the meter stops; end the session and the machine is wiped.
Buy a server per person, or rent one by the hour.
A bought server runs, and bills you, around the clock. The average employee who actively uses AI touches it just 1 to 1.5 hours a day. You'd be paying for the other ~85%.
Drag the sliders below to see your team's real cost either way.
The average across companies
1–1.5hours a day
Active AI use per employee: about 5–6 hours a week, or 15–20% of a 40-hour week.
A bought server bills you for every hour of every day. Rent and you pay for the ~6 hours your team needs, not the idle ~34.
Buy a server each · 3 years
€69,000
€60,000 upfront + power · 1 user each
Pryvan · 3 years
€11,880
≈ €330/mo · €0 upfront · unlimited at once
You save · 3 years
€57,120
vs buying 10 servers
Cumulative cost over 3 years
Connect to a server. Log off. Everything's wiped.
Pryvan runs on hundreds of AI servers across the world, always on and ready. You connect to one that's already running, do your AI work, then log off — and the moment you do, everything on that server is deleted. You're offered a download of your output (or the whole chat) to keep on your own computer or phone. The minute-meter stops the instant you log off, and your prepaid balance drops only by the minutes you actually used.
Hundreds of servers worldwide
Always-on AI servers — connect to one that's already running, no waiting.
Anonymous & hyper-secure
No profile attached to a session. Just connect, work, and log off.
Wiped when you log off
Every session's data is deleted on log-off — download your output first if you want to keep it.
Pay by the minute
The meter stops the second you log off; your prepaid balance drops only by minutes used.
Pick the model. We pick the machine.
From everyday assistants to the strongest open models for coding, vision and reasoning — choose what your task needs and Pryvan runs it on a fresh machine for you.
| Task | Recommended model | Tier |
|---|---|---|
| Coding | Qwen2.5-CoderTop open-weights coder, with full-repo context on Max. | Core + Max |
| Agents & general chat | Llama 4 ScoutFast, multimodal and very long context. | Max |
| Deep reasoning | DeepSeek-R1Step-by-step reasoning for analysis and maths. | Max |
| Long documents | Llama 4 ScoutReads book-length context in one go. | Max |
| Vision & scanned docs | Qwen2.5-VLReads PDFs, screenshots, photos and tables. | Max |
| Max-quality frontier | Qwen3 (largest)The strongest open model on the network. | Max |
| Everyday assistant | Gemma 3 · Mistral SmallFast drafting, summaries and Q&A. | Core + Max |
| Fast & cheap | Qwen3 (MoE)Very responsive on the Core machine. | Core + Max |
No subscriptions. No surprises.
No monthly fee, no per-token metering, no charge for idle time. You only pay for compute you actually use.
- Idle time — billing stops when the machine is off
- Per-token or per-query metering
- Legal documents & AI-Act risk classification
- Setup or onboarding fees
Questions about the network or volume? Read the FAQ talk to us.
Bring AI into your business, without giving up your data.
Join the waitlist. We're onboarding GDPR-sensitive SMEs across Europe.