Documentation
Everything you need to understand and use the GPUVerse infrastructure decision engine.
Introduction
GPUVerse is the intelligence layer above every cloud. It answers a single hard question for AI teams: where and how should this workload run? Discover, the first product, turns a workload description into an explainable infrastructure plan.
Core concepts
A Workload is what you want to run (model, size, scale, latency, budget, compliance). A Recommendation is GPUVerse's explainable answer - provider, GPU, region, cost, tradeoffs, and alternatives. A Plan is a saved, versioned recommendation you can export and deploy.
The Discover workflow
Submit a workload and GPUVerse reasons through six steps: understanding the workload, matching model requirements, checking GPU options, comparing providers, estimating cost, and generating the plan. The output is structured - never a black-box paragraph.
Providers & GPUs
GPUVerse scores 10 providers across hyperscalers, neoclouds, and marketplaces, and 8 accelerators from H100 to TPU v5e. Each is normalized on price, availability, region coverage, compliance, and reliability.
API reference
The GPUVerse API returns structured recommendations designed to drive provisioning. The endpoint below is illustrative - the public API launches with Provision.
POST https://api.gpuverse.ai/v1/discover
{
"workload_type": "inference",
"model": "Llama 3.1",
"model_size": "34B-70B",
"scale": "large",
"latency": "interactive",
"region": "eu-west",
"budget": 3000,
"compliance": ["SOC 2"],
"goal": "balanced"
}Ready to design your infrastructure?
Open Discover and get a plan in seconds.