Novita AI (United States) is an AI cloud aimed at developers and agents. The company serves models through a token-billed serverless API and offers dedicated GPU endpoints for your own models or adapters. The same account also reaches GPU instances and an isolated environment for running agent-generated code.
What it does
- Consumes text, image, audio and video models through a single API
- Publishes your own models or LoRA adapters to dedicated endpoints
- Prototypes with starter credits and zero-cost models before scaling
How it works
- The serverless API is compatible with OpenAI ecosystem clients and bills per token consumed
- Dedicated endpoints reserve a GPU for a period and accept your own weights
- The catalog gathers open models from several vendors, including image and video generation
- Batch inference and isolated execution of agent code are supported
Strengths
- Per-token billing in serverless mode, with no idle GPU rental
- Self-service entry with trial credits and zero-cost models
- One platform covers text, image, audio and video
Points of attention
- The company also rents GPUs by the hour; this card covers the model API
- Frontier models may differ in price and availability from the rest of the catalog
- Enterprise features and higher limits go through sales
