Your AI,
Your rules
Dedicated GPU instances with open-source LLMs. Documents never leave. Self-destructs when done.
1. Non-solicitation clause extends 24 months post-termination, covering all subsidiaries
2. Liquidated damages set at $500K per breach, enforceable without proof of actual loss
3. Governing law defaults to Delaware courts with mandatory arbitration waiver
What you get
Security By Architecture,
Not By Promise
Dedicated GPU
Your own isolated GPU instance. No shared resources, no noisy neighbors, no compromises.
Self-Destructing Vaults
Workspace and everything in it is destroyed when you are done, and cannot be restored.
Open Source Models
Fully auditable LLMs. No black boxes.
Full Audit Trail
Every action timestamped — never your documents or answers.
Zero Knowledge
We never see your documents. Ever.
Network Isolation
Your vault runs on hardware rented for you alone. No third-party AI API sees your documents.
Simple pricing
Pay As
You Go
No subscriptions. Top up any amount from €1 to €20,000, or its equivalent in the currency you pick at the day's rate. Spin up a vault, use it, destroy it. You only pay for GPU time consumed.
$0.99/hr
Starting rate · metered by the second · $0.0165/min
FLASH
Qwen 3 — 4B
$0.99/hr
$0.0165/min
TURBO
Qwen 3.5 — 35B MoE
$4.99/hr
$0.0832/min
ULTRAComing Soon
GLM-5 — 744B MoE
FAQ
Frequently asked questions
Everything you need to know about VaultAI's security and privacy architecture.
ChatGPT processes your documents on shared infrastructure, and your data may be used for training. VaultAI spins up a completely isolated GPU instance for each vault. Your documents never leave your dedicated environment, are never shared, and are never used for model training. When you destroy the vault, the data is permanently deleted.
Your documents travel over HTTPS (TLS) to your dedicated GPU instance. This is not end-to-end encryption: the instance reads your documents in order to index them. They are processed locally on that instance using open-source AI models. Our orchestration layer only handles metadata (vault status, billing) and never has access to your actual document content.
When you destroy a vault, the dedicated GPU instance is completely terminated. All data, including uploaded documents, AI model memory, chat history, and generated outputs, is permanently deleted. The underlying storage is wiped and the instance is decommissioned. This process is irreversible by design. What remains is the audit trail for your account: a timestamped record that the vault was created, used and destroyed — the actions, never the content of your documents, questions or answers.
We run open-source models: Qwen 3 4B on FLASH and Qwen 3.5 35B MoE on TURBO, with GLM-5 744B MoE on ULTRA coming soon. They are served via vLLM, the fastest open-source inference engine. All model weights are fully auditable and run entirely on your dedicated GPU instance — no data is sent to any third-party API. You choose the model tier that fits your workload, from fast extraction to full-precision reasoning.
VaultAI is not HIPAA certified, and a Business Associate Agreement (BAA) is not currently part of the service. Each vault runs on a GPU instance rented for you alone, and the instance and its data are deleted when you destroy the vault. If your work involves protected health information that requires a BAA, please do not upload it to VaultAI.
VaultAI uses simple pay-as-you-go pricing with no subscriptions. Each top-up is any amount from €1 to €20,000, or its equivalent in the currency you pick at the day's rate. GPU time starts at $0.99/hr for the FLASH tier (4B model) and $4.99/hr for the TURBO tier (35B MoE). The ULTRA tier (744B MoE frontier model, 40B active) is coming soon and cannot be ordered yet. You only pay for the minutes you use. Add credit to your balance and start analyzing — no commitments.