LLM inference
Serve or evaluate language models with the memory and context they require.
Hostnot GPU / infrastructure
GPU cloud compute for AI, inference and development
Hostnot GPU provides on-demand GPU instances, supported serverless inference, and developer APIs with customer-facing pricing and wallet authorization.
Compute without the cloud theater. GPU compute, serverless inference, and precise wallet billing from hostnotgpu.
SERVER-SIDE PRICING · API READY
Control planereal capacity, one workspace
187Available offerings
27GPU types
8Regions
49Serverless models
BUILT FOR SHIPPING
Infrastructure that stays legible under load.
Operational detail where you need it, without hiding the important decisions behind a checkout flow.
01Live GPU availabilityBrowse customer-visible capacity before you deploy.↗02Wallet-first authorizationReview the approved rate before compute is requested.↗03SSH-ready instancesBring a public key and keep private access private.↗04Serverless inferenceDiscover supported models through a customer-safe API.↗05Deploy templatesStart from a repeatable environment for your workload.↗06Scoped developer APIAutomate capacity, usage, and inference with clear boundaries.↗
GPU CLOUD / CAPACITY
Choose compute that matches the workload.
Review live GPU capacity by model, memory, region, and capacity class instead of choosing from a static inventory list.
Use an instance for SSH access, custom environments, persistent experiments, fine-tuning, rendering, or a self-managed service.
Read the GPU Cloud guideMARKETPLACE / LIVE CATALOG
Rates and availability are read from the customer-visible catalog.
serverless request
POST serverless.hostnotgpu.ae/v1/inference/chat
Authorization: Bearer YOUR_API_KEY
Idempotency-Key: request-unique-001
{ "model": "MODEL_SLUG",
"messages": [{"role": "user",
"content": "Hello"}] }
200 / inference response validated
SERVERLESS AI / API
Deploy inference through a customer-safe API.
Discover the current model catalog, validate capabilities, and send chat, image, or video requests with scoped credentials and unique idempotency keys.
Wallet authorization and usage visibility are handled server-side. Provider credentials and internal routing stay behind the platform boundary.
Open the Serverless API guideHOW IT WORKS
Prepare, connect, and complete your first GPU workload.
Choose a compatible configuration before authorizing compute, then verify access and track the instance through completion.
01PREPARE & AUTHORIZE
Create an account, fund your wallet, and register a public SSH key. In the Marketplace, review GPU count, memory, storage, region, capacity class, and compatible environment. Confirm the authorized rate before requesting deployment; available wallet balance excludes active reservations.
Open guide →02CONNECT & VERIFYWait for provisioning and the displayed connection details. Use the instance's SSH host, user, and port, and verify the displayed SHA256 fingerprint. Keep the private key on your device. If access fails, capture the instance status and redacted error before troubleshooting.
Open guide →03FINISH & REVIEWSave the workload outputs you need before terminating an instance. Closing a terminal is not an instance lifecycle action: use the workspace controls and check the resulting status. Review usage and transactions for completed billing windows after the run.
Open guide →WORKLOADS
Built for practical AI infrastructure.
Choose a runtime that fits the work, then keep the artifacts, credentials, and operational controls clear.
Fine-tuning
Run controlled experiments with checkpoints and repeatable environments.
AI development
Build notebooks, APIs, and prototypes with direct accelerator access.
Computer vision
Process image and video workloads on appropriately sized GPUs.
Rendering
Use GPU compute for visual workloads and batch processing.
Custom environments
Bring compatible dependencies, containers, and model artifacts.
FAQ / BEFORE YOU DEPLOY
Everything you need before deploying.
Short answers with links to the deeper operational guides.
What is Hostnot GPU?+
Hostnot GPU provides customer-facing GPU cloud compute, serverless inference, and developer APIs through one workspace. The public Marketplace and model catalog show the current customer-visible options.
How do I choose the right GPU?+
Start with the live GPU Marketplace and compare memory, GPU count, region, environment, capacity class, and displayed customer rate. Match those details to the model or workload instead of choosing by GPU name alone.
View current GPU capacity →What is the difference between GPU Instances and Serverless AI?+
GPU Instances provide a persistent environment with SSH access and control over your runtime. Serverless AI is for supported request and response workflows where the current model catalog and API contract fit your application.
Compare Serverless AI →How does GPU pricing work?+
Customer rates are calculated server-side from the selected offering and current platform configuration. Review the displayed rate before deployment; availability and rates can change as capacity synchronizes.
How do I connect to a GPU instance?+
Register a public SSH key, wait for the instance connection details, verify the host fingerprint, and connect using the host and port shown in the workspace. Never send your private key to support.
Read the SSH guide →Do I need an SSH key?+
You need an approved access method for the instance. SSH public keys are the normal workflow for Linux GPU instances because the private key stays on your device and is not uploaded to the platform.
Can I use Hostnot GPU through an API?+
Yes. The developer API exposes customer-scoped models, instances, usage, and wallet-aware operations. Serverless requests use scoped API keys, capability validation, and unique idempotency keys.
Open API documentation →How does billing work?+
Deployments use wallet authorization before a compute resource is requested. Usage is recorded after the applicable billable state or request outcome, and account-specific records remain visible in the authenticated workspace.
Can I run open-source AI models?+
Custom model workloads can run on GPU Instances when the selected environment and hardware support them. Check the model license, runtime requirements, memory estimate, and deployment documentation before starting.
Where can I see current GPU availability?+
The GPU Marketplace is the source for current customer-visible capacity, configurations, regions, and rates. Static articles should not be treated as inventory promises.
How do I get technical support?+
Use the public Support form or the Support Inbox in your workspace. Include safe identifiers, timestamps, and redacted errors. Never share passwords, private SSH keys, API secrets, or payment credentials.
Contact Support →EXPLORE
Start with the workflow you need.
Clear paths from first deployment to production operations.
01GPU QuickstartLaunch and connect to your first GPU instance.↗02Serverless APIMake a capability-gated inference request.↗03SSH AccessGenerate a key and connect safely.↗04TroubleshootingDiagnose common connection and workload issues.↗05Developer SecurityProtect API and SSH access from day one.↗06Technical GuidesDeep GPU and AI tutorials for builders.↗
READY TO BUILD?
Deploy when the workload is ready.
Browse current capacity, read the docs, and choose the path that fits your next system.