Novita AI: AI cloud for developers: model APIs and GPUs

What Novita AI is

Novita AI describes itself as an AI-native cloud for builders and agents. It combines model APIs, GPU infrastructure and agent runtimes in one platform.

Novita AI key features

• Model APIs: serverless access to 200+ models across text, image, audio and video, billed per usage
• Dedicated Endpoints: private, isolated endpoints for your own model and compute
• Agent Sandbox: isolated runtimes for agents that run tools and execute tasks, billed per second
• GPU Instances: dedicated GPU machines for deploying models, inference and training
• Serverless GPU: submit jobs without provisioning instances, with scaling down to zero
• Bare Metal: dedicated GPU clusters, including 8x NVIDIA H200 and 8x NVIDIA H100 configurations
• Partner listings include Hugging Face and OpenRouter, and the site shows a SOC 2 badge

Try Novita AI on the official website

Novita AI pricing at a glance

• Model APIs are billed by usage: per million tokens for language models, per image or per second for image and video, and per request for search.
• Examples from the pricing page: DeepSeek V4 Flash at $0.14 input and $0.28 output per million tokens; Llama 3.1 8B Instruct at $0.02 input and $0.05 output; Qwen-Image text to image at $0.02 per image.
• Batch inference has an introductory 50% discount on input and output tokens for supported models.
• Plan tiers, billing terms and refund terms were not shown on the pricing page. Dedicated Endpoints, Agent Sandbox and GPU prices are on separate tabs.

Free plan or trial: The website says Novita AI is free to start. Free credit amounts and trial terms were not stated, and some models are listed as free.

Prices change often. Confirm on the official website before you buy.

Who Novita AI suits

Developers, AI agent builders and teams that want model APIs, GPU instances or dedicated clusters without running their own infrastructure.

Novita AI strengths and limits

More about Novita AI