Use the current displayed rate for your GPU count, duration and start time. Prices can change as available capacity changes; a confirmed reservation keeps its accepted rate.
Read the price
Rates are shown per GPU-hour. The quote shows the total for the selected GPU count and duration, with the tax treatment identified before confirmation. Business rates marked excluding VAT are not tax-inclusive consumer prices.
There is no purchasable rate when no matching offer is available. Submit a request to discuss your requirements instead.
Compare start times
The market chart shows capacity available for each complete window. Selecting a start time shows the current offer for that window, rather than an estimate of what its price might become later.
A larger available GPU count does not guarantee that every requested configuration fits: model, location, topology and supplier conditions also matter.
Accepted terms
Your confirmation records the agreed price and terms. Review the quote deadline and cancellation conditions before accepting. Cancellation and release costs depend on the reservation's accepted terms, not the current market rate.
For help choosing a configuration, submit your workload.
Disk and transfers
Your final quote separates GPU and included resources from additional working disk. Select the disk before reserving: the available choices depend on the supplier, from 10 to 2,000 GiB. The supplier specifies the included disk and maximum capacity. CPU and RAM remain included at 4 vCPU and 16 GiB per GPU.
Additional disk is priced per 100 GiB-hour and charged for delivered session minutes, with the line rounded once to cents. It is not multiplied by the GPU count. For example, 250 GiB selected, 100 GiB included and €0.20/100 GiB-hour adds €0.30 for one delivered hour. GPU and disk credits are held together before reservation. Provisioning without hand-over incurs no usage charge. Cancellation follows the accepted cancellation schedule.
Upload to the GPU and download from the GPU have separate included allowances and extension rates. The workspace displays both remaining budgets and warnings at 80% allocated. SSH, SFTP, VS Code and loopback forwarding use these budgets. Both directions need remaining capacity for SSH to work. Compute continues until stopped or expired if transfers are blocked; keep download capacity for your results.
Disk is temporary and fixed for the session. Kubernetes reserves the selected size plus runtime overhead. Exceeding the disk limit can evict the workload; this is not a synchronous write quota. Files are deleted at cleanup. Download or export important results before the reservation ends.
Earlier accepted profiles keep their 10 GiB disk and combined transfer allowance; they do not acquire new prices or limits.
Prepaid transfer extensions
The supplier fixes separate upload and download rates per 100 GiB, which are copied into the accepted quote. Choose additional capacity in 100 GiB packs and explicitly confirm its maximum total. The credit hold and quota increase happen together. A €0 rate adds capacity free of charge, still within the accepted maximum. There is no automatic renewal or postpaid overage invoice.
After the included directional allowance is exhausted, each started GiB of reserved transport capacity is charged. The final amount is calculated from the accepted 100 GiB rate and rounded once per extension. Unused whole GiB return to your credits when the session ends. For example, a 100 GiB extension at €2.00/100 GiB holds €2.00. If 12.2 GiB are allocated, 13 GiB cost €0.26 and €1.74 returns to credits.
The relay allocates blocks of up to 1 MiB before forwarding. SSH overhead and unused block tails remain included after disconnects. This is a reserved-transport counter, not a meter of exact file sizes or the cloud provider’s invoice. The direction is determined by the relay, not declared by customer software.
The reservation shows the total authorised for GPU, disk and purchased transfers, plus extension holds, charges and returned credits. Check purchase history if a response is interrupted. Buying transfer does not extend GPU time or disk size. A full service reversal reverses the resource charges too. Historical per-GiB extension contracts retain their accepted quantities and rates.
Cloud providers and personal buckets
Supplier prices reflect the specific compute contract, region, storage and network costs. No public provider rate or shared promotional allowance is applied automatically. A provider’s Object Storage tariff is not necessarily its GPU instance’s network tariff. Supplier image downloads and infrastructure overhead must be covered by its accepted rates; they cannot generate an unapproved customer invoice.
A personal S3 bucket is billed directly by its provider. Storage, requests, retrieval and outgoing traffic on that account are outside the Requenta total. A transfer from a personal bucket to the GPU can incur bucket-side outgoing fees and use the Requenta upload allowance. Exporting GPU results uses the download allowance. Check both ends of the route; the same service must not be charged twice by Requenta.
Direct Internet access from the GPU container is disabled. Use SSH/SFTP to transfer files, including files fetched from a bucket on your computer. A saved S3 address does not connect or mount a bucket. See Workspaces.
Supplier earnings
New reservations allocate 94.5% to the supplier and 5.5% to Requenta before costs. No shared bonus is accrued. GPU and disk settle together; each transfer extension settles separately. Supplier earnings are rounded once to cents, with the commission as the remainder. Historical accepted settlement terms remain unchanged.
The supplier form accepts net resource rates and displays the corresponding customer price. Rates use a 100 GiB basis to represent small per-GiB amounts without floating-point arithmetic. Settlement details show actual supplier earnings and commission after usage ends.