About Novita.ai
Novita AI is a cloud platform for developers that bundles model APIs, agent sandboxes and GPU instances. The vendor describes it as an AI native cloud for builders and agents, with access to more than 200 models through a single API.
Model APIs are serverless and cover language, image, audio, video and vision models, billed by the token rather than by the hour. Listed examples include DeepSeek, Kimi, GLM, MiniMax, Qwen and Gemma families, and batch inference is offered at an introductory 50 percent discount on supported models. Dedicated endpoints give private, isolated capacity for consistent latency. Agent Sandbox provides isolated runtimes for agents that run code and call tools, billed per second with a stated startup of about 200 milliseconds. GPU Cloud offers instances for training and inference. The platform is aimed at developers and startups building AI products.
Pricing is published per million tokens in dollars. On the pricing page, DeepSeek V4 Pro lists $1.6 input and $3.2 output, DeepSeek V4.1 Flash lists $0.3 input and $1.2 output, and DeepSeek R1 0528 lists $0.7 input and $2.5 output, with cache read rates shown for some models. Sandbox and GPU rates are published on separate tabs.
The home page carries copy addressed to AI agents, including an instruction to read a skill file and follow it.