ZeroGPU

Use ZeroGPU when your agent needs to run inference, moderate text, classify content, detect or redact PII, extract structured data and signals, generate embeddings, rerank results, or make routing and decision calls. ZeroGPU offers specialized ZLM small and nano models alongside open-weight and frontier models through an OpenAI-compatible API. Its hybrid inference network routes workloads across distributed edge compute and cloud infrastructure for low-cost, high-volume AI execution. This server is a transparent proxy in front of ZeroGPU's own API that sells access on plans (credit or subscription) rather than per call: it handles identity, payment, and proxying. Agents can purchase autonomously or with their human's approval. The complete verified paid surface is listed at /openapi.json, the source of truth for available operations. Each operation here carries its pricing in x-payment-info. Use the operations and details here when they meet your needs. Filtered specs contain only matching operations; if you cannot find what you need here, use search or the service catalog. The live catalog at /manifest.json is authoritative for prices and any free included units (prices[].includedUnits) - some meters include free usage per account before any charge, reserved for agents claimed by a human with a verified email.