Products
Use compute when the work needs it.
Use ICPX for agent runs, model deploys, startup ML experiments and private GPU sessions. Keep every run scoped, priced and easy to close.
Create accountICPX GPU cloud
Compute on demand.
Run on-demand compute for AI agents, inference, fine-tuning and evaluation. Search models, deploy a run and keep terminal access scoped to the workload.
AI agent workloads
One-click deploys
ML for startups
AI agent runs
Give agents GPU tools, logs and a clean runtime for research, automation and evaluation.
AgentsToolsLogs
One-click deploys
Start from a model or skill, pick capacity and launch a run without building cloud glue.
ModelsSkillsDeploy
Machine learning for startups
Prototype, fine-tune and validate models without buying hardware or reserving clusters.
Fine-tuneBatchEval
Private GPU sessions
Open SSH, stream logs and close the job when the workload is complete.
SSHTerminalReceipts
Choose GPUs
Select model family, VRAM class, region and max spend before launch.
Deploy models
Search NVIDIA models and skills, then launch the workload from one flow.
Run agents
Give agents isolated compute, terminal access and live logs per job.
Train and evaluate
Run fine-tunes, batch inference, validation sets and model comparisons.
Metered runs
Fund compute with a wallet, cap spend and refund unused balance at close.
Secure access
Use MFA, CSRF controls, signed wallet actions and scoped SSH keys.
Model index
Search the model. Launch the job.
Start from a model, skill or blueprint, then pick the GPU class that fits the workload.
0
Models
0
Skills
0
Blueprints
0
GPUs
No matching models. Try a model family, task type or runtime name.
How it works
From model idea to paid compute job.
ICPX keeps the path short. Search, rent, run and settle without switching between cloud consoles, wallet tools and terminal scripts.
Search
Find a model, skill or workload template that matches the job.
Rent
Choose GPU capacity and approve the quote before anything runs.
Run
Open logs, SSH or terminal access while the job is active.
Settle
Pay for completed compute and return unused balance to the wallet.
Wallet payment
Pay for compute when the job runs.
Connect Phantom, Solflare or Fuse, fund the run and settle usage from the wallet attached to the account.
Agent workloads
Long-running tool use, browser tasks and evaluation loops
Model deploys
Inference, embeddings, voice, vision and reranking jobs
Startup ML
Fine-tuning, dataset tests, demos and product validation
Provider capacity
GPU supply, usage receipts, health and settlement
Compute payment
Pay for compute with ICPX.
Use ICPX to fund compute jobs, set spend caps, release provider payments and refund unused balances when the job closes.
Compute
Usage
1,000,000,000
Supply
DexscreenerContract
HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump
Start