RTX 4090
24 GBExperiments & inference
Memory / GPUChoose a GPU model and capacity for your workload.
Open a model to review capacity, environment and order details.
Experiments & inference
Memory / GPULarge-memory workloads
Memory / GPULarge-model training & inference
Memory / GPUTraining & large models
Memory / GPUCompare model, memory, CPU, RAM, SSD and hourly price side by side.
Ubuntu, JupyterLab, VS Code, PyTorch, vLLM, ComfyUI, Text Generation WebUI or your own image.
Prepaid balance with per-second billing; you only pay while the instance runs. Listed prices include 5% business tax.
Planned services for laboratories, schools and companies.
Buyer name and tax ID can be confirmed before purchase. Choose an e-invoice or a paper handwritten three-part invoice.
Quotations and supporting documents are prepared around your institution's reimbursement rules.
Test a workload first with free research GPU credit, usable on any configuration.
Invite lab or team members, share usage and keep billing in one place.
This applies once the service opens; reimbursement remains subject to your institution's rules.
Answers reflect the current state of the service.
Not yet. ZaoSeq currently accepts requirement and partnership inquiries only. Self-service rental and payment are not open, and the configurations in the catalog are examples of how ordering will work.
A prepaid balance billed per second while an instance runs. Listed prices include 5% business tax, and the hourly price is shown before you launch.
Yes. Once the service opens, e-invoices are sent by email free of charge, with no order minimum. You can also choose a paper handwritten three-part invoice, sent by registered mail within 5 working days of confirmed payment; postage is free for orders of NT$3,000 or more (tax included) and a NT$35 delivery fee applies otherwise.
Buyer name, tax ID and supporting documents can be discussed before purchase. Whether a purchase can be reimbursed depends on your institution's rules, so please confirm them first.
RTX 4090, RTX PRO 6000, H200 and B200, in 1, 2, 8 or 16 GPU configurations. Clusters of more than 16 GPUs are handled as project inquiries.
Machines are supplied by partner data centers rather than ZaoSeq-owned facilities. The region depends on the configuration or project and is shown when you choose a configuration. Please do not send sensitive data through the inquiry form.
Service, trial, discounts and invoicing are being prepared; current supply cards are examples, not available inventory or binding offers.