Platform Overview
This platform provides AI cloud services. You can call the published AI models directly, deploy your own private model endpoints, or rent GPU containers and storage to run your own workloads — all billed by actual usage, so you only pay for what you use.
Three ways to use the platform
Call public model APIs (billed per token) No infrastructure to manage. Pick a model from the Models Hub, get an API key, and call it through an OpenAI-compatible API, billed by token usage. Ideal for developers who want to integrate AI capabilities into their applications quickly.
Deploy private model endpoints (billed by resource hours) Deploy a model as your own dedicated inference endpoint with exclusive compute resources and auto scaling (including Scale-to-Zero). Ideal for production services with latency, throughput, or data-isolation requirements.
Rent GPU containers (billed by resource hours) Rent GPU compute as Pods, using official image templates or your own custom images. You can mount persistent storage and work in the container directly through the Web Terminal. Ideal for training, experiments, and batch jobs.
Recommended reading path
- Sign up and log in — create an account
- Quickstart: call your first model API — complete your first inference call in 5 minutes
- Quickstart: deploy your first Pod — rent a GPU container
- Core concepts and terminology — understand the platform's basic objects
Depending on your development scenario, you can also start directly from the examples.