Shared GPU

Your private AI on a real GPU, without server administration

An OpenAI-compatible endpoint with OCR, Whisper and voice already installed. Access is controlled by token and IP addresses, with at most 2 tenants per GPU, no content recorded and service ready within 48 h.

Up to 2 tenants
Paano ito gumagana

Connect your application. We take care of the GPU.

OpenAI-compatible endpoint

Connect your integration using the URL and model shown in the customer portal. You do not have access to the machine.

Token and registered IPs

We accept only requests that include a valid token and come from an allowed IP address.

At most 2 tenants per GPU

Each tenant has their own concurrency slots and context limits. Performance may vary with the other tenant’s usage.

Private content

Request and response content is never stored. Only usage counts and technical metadata are kept.

It comes with OCR, Whisper (transcription) and voice through the same endpoint.

Waitlist

Choose your private AI

Monthly reference price in US$. Billed in your account currency.

Coming soon

New GPUs for private AI are on the way

We're preparing our own cards for the Shared GPU service, with OCR, Whisper and voice, plus 10 GB of storage per tenant. Leave your contact details and we'll notify you when the service opens, with no obligation. Pricing and the launch date are only confirmed once the card is listed.

Ready-to-use private AI

Ready-to-use private AI goes beyond a low-cost GPU slice.

An OpenAI-compatible endpoint with OCR, Whisper and voice already installed. Access is controlled by token and IP addresses, each GPU has at most 2 tenants, and no content is recorded. Ready within 48 h, with no server to manage.

What’s included

Ready-to-use endpoint

OpenAI-compatible API with OCR, Whisper and voice installed and maintained by GeHost.

Monitored service

A monitor checks the health of the service on the GPU and restarts it automatically if it stops.

Privacy by default

Access via token and registered IPs. We do not log any request or response content.

Share the GPU and pay less

The listed price applies when you have the GPU to yourself. When a second tenant joins, the discount is automatically applied to both invoices.

GPU replacement at no extra cost

If the GPU becomes unavailable, GeHost migrates the service to an equivalent one.

Ready within 48 h

After payment, the endpoint, token and IP addresses are available in the client area.

Comparison

Building and maintaining your own AI vs. GeHost Shared GPU

Build it yourselfGeHost
ServerRent the machine and install drivers, models and an inference serverReady-to-use OpenAI-compatible endpoint with OCR, Whisper and voice
MaintenanceYou monitor, restart, update and replace the machineGeHost monitors and restarts the service and replaces the GPU at no extra cost
SecurityYou configure the token, firewall and loggingToken and registered IPs. No content logging.
CostPay for the whole GPU, even with low usageUp to 2 tenants, with an automatic invoice discount when the second joins
SuportaKnowledge base, forum or ticket in English24/7 AI with immediate responses in Portuguese, Japanese and English
Who it’s for

Built for customers who need responsive operations

OCR and documents

Read PDFs and images with private AI without sending data to public services.

Transcription and voice

Whisper and voice through the same endpoint for customer support, meetings and media.

Agencies and integrators

Private AI for your clients’ projects, with no server to set up or monitor.

Operations in Brazil and Japan

Support in Portuguese and Japanese and billing in Brazilian real or Japanese yen.

Contact GeHost before you order

Tell us about your workload and timeline. We respond in Portuguese, English or Japanese.

Which GPU should you choose?

Ready-to-use AI or an exclusive GPU

Shared GPU

Endpoint and model prepared by GeHost, with up to two tenants. Use AI without administering the server.


Dedicated GPU

For those who need exclusivity and administrative access to the server.

Explore Dedicated GPU
Mga madalas itanong

Before you order

No. SSH and sudo are unavailable. Access is limited to the endpoint and requires a token and a registered IP address. GeHost manages the machine.

Each tenant has their own storage with a quota. Temporary files are deleted after 24 h; other files remain until you delete them or terminate the service. Conversations are never stored.

OCR, Whisper (transcription) and voice through the same OpenAI-compatible endpoint.

Once two paying tenants share the GPU, the shared GPU discount applies to each new invoice. We notify you by email of changes in the number of tenants before your next invoice. Internal use does not count.

Yes. The GPU serves up to two tenants and performance may vary with the other tenant’s usage. For exclusive use, choose Dedicated GPU.

Within 48 h after payment confirmation. Delivery may involve replacing the GPU with an equivalent one.

No. Content is never stored; we keep only usage counts and technical metadata.

Individuals can cancel for a full refund within 7 days of signing up. After that, canceling does not entitle you to a refund for the current month, and service continues until the end of the paid period.

GeHost migrates the service to another equivalent GPU at no cost. If none is available, you can cancel with a prorated refund.

Everything on the machine: the GPU, system, OCR, Whisper, voice, inference server, health monitoring and GPU replacement. You get the endpoint, token and IP address controls in the client area.

With Shared GPU, you use ready-to-run private AI without access to the machine and share the GPU with no more than one other tenant. With Dedicated GPU, the entire card is yours, you have root access and you choose what runs on it.

GeHost's AI responds immediately in Portuguese, Japanese and English, 24 hours a day, and also monitors the infrastructure for your service.

Your application, your private AI

Join the waitlist or contact GeHost to size your setup for your workload.

Join the waitlistMakipag-usap sa GeHost

See the Terms of Service, section 8.5.