Novita: Run models and GPUs on one platform
Novita is an AI cloud platform that offers serverless model APIs, dedicated endpoints, GPU instances and agent sandboxes for developers.
What is Novita?
Novita is an AI cloud aimed at developers and agents. It combines serverless model APIs covering text, image, audio, video and vision models, dedicated private endpoints, GPU instances, serverless GPU jobs and bare metal clusters, plus an agent sandbox for isolated agent runtimes. The platform states 200+ models, 200ms latency and 99.5% uptime, and lists per-model token rates such as DeepSeek V4 Pro at $1.6/Mt input and $3.2/Mt output. The GPU side covers dedicated instances, serverless jobs that scale to zero, and bare metal clusters with NVIDIA H200 and H100 hardware on NVLink and RDMA interconnects. Agent sandboxes bill per second and are described as isolated runtimes where agents run tools and execute tasks. Novita says it is free to start and scales as workloads grow, with batch inference available at an introductory 50% discount on input and output tokens for supported models.
Focus areas: Model API.
Who is Novita for?
- Developers building AI agents that need isolated runtimes
- Engineering teams running inference without managing servers
- Startups that need GPU capacity for training or fine-tuning
Is Novita worth it?
A broad AI infrastructure platform that keeps model APIs, GPU capacity and agent sandboxes in one place for developer teams
Where Novita is strong
- Model APIs, GPU compute and agent runtimes sit on one platform
- Serverless options bill by token or execution rather than idle time
- Per-model token pricing is published on the pricing page
Where it falls short
- No fixed subscription tiers or plan names are listed, only usage rates
- The submitted description mentions image generation for blogs, which does not match the developer-focused platform content
How do you get started with Novita?
- 01Create an account and get an API key from the Novita dashboard
- 02Pick a model from the API catalogue and call it through the single endpoint
- 03Provision a GPU instance, serverless job or agent sandbox when you need compute
Novita: frequently asked questions
What can I run on Novita?
The platform offers serverless model APIs for text, image, audio, video and vision, dedicated private endpoints, GPU instances, serverless GPU jobs, bare metal clusters and agent sandboxes.
How is Novita priced?
Model APIs are billed by the token with per-model input, output and cache-read rates listed on the pricing page, while agent sandboxes bill per second and serverless GPU jobs bill for execution. Batch inference is offered at an introductory 50% discount on input and output tokens for supported models.
How many models are available?
Novita states that 200+ models are available through a single API, including DeepSeek, GLM, Kimi, Llama and Qwen families with context windows up to 1M tokens.
What hardware do the GPU offerings use?
The GPU cloud section lists clusters with 8x NVIDIA H200 (141 GB HBM3e per GPU) and 8x NVIDIA H100 (80 GB HBM3 per GPU) nodes, connected with NVLink 4th Gen and 400 Gb/s RDMA networking.
Novita alternatives
The closest tools in our directory, matched on category, focus areas and audience.
| Tool | Best for | Free plan | Starting price | Focus areas |
|---|---|---|---|---|
| Novita | Developers running model inference, GPU jobs and agent workloads | No | — | Model API |
| iCreat AI | ICreat AI - Your All-in-One AI Image & Video Creator, Create More, Spend Less. | Unknown | — | Model API |
| Synexa | Synexa AI: Deploy AI models with one line of code, cost-effective, scalable GPU infrastructure. | Unknown | — | Model API |
| Chutes.ai | Developers running open-source model inference at scale | No | — | Model API, Model Inference |
| ModelsLab | Developers who want many generative models behind one API | No | $21/mo | — |
| PeriFlow | Teams serving open-weight models to production agents | No | — | Model API, Model Inference |
| Runware | Developers running many AI models or custom GPU workloads | No | — | — |
| Flaq AI | Developers routing image, video and LLM calls through one API | No | — | — |
| APIMart | Developers who want one API key across many AI models | No | — | — |
| PoYo.ai | Developers needing many generation models behind one API | No | — | — |
| Cloudglue | Developers adding video search and chat to AI products | Yes | $15/mo | — |
