AI cloud for developers
Novita AI Pricing
Novita AI charges for what you use: per token for model APIs, per second for the sandbox and per hour for GPUs. This page lists the prices from the official site.
See Novita AI PricingSee The Prices
Affiliate link: opens the official website. See the disclosure below.
Model API Prices Per Token
Prices below are as listed on the Novita AI pricing and GPU pages on 24 October 2026. The home page showed a different figure for DeepSeek V4 Pro, so treat every number as a snapshot and check the live page.
| Model | Input, per million tokens | Output, per million tokens |
|---|---|---|
| DeepSeek V4 Pro | $1.6 | $3.2 |
| DeepSeek V4.1 Flash | $0.3 | $1.2 |
| GLM 5.3 | $1.4 | $4.4 |
| Kimi K2.6 | $0.8 | $3.4 |
| Qwen3.8 Max | $2 | $6 |
| Llama 3.3 70B Instruct | $0.135 | $0.4 |
| OpenAI GPT OSS 120B | $0.05 | $0.25 |
The pricing page also lists cache-read prices for some models, embeddings (Qwen3 Embedding 8B at $0.07 and BAAI BGE-M3 at $0.01 per million tokens) and AI search calls ($0.007 to $0.016 per request, depending on the search type).
GPU Hourly Prices
| GPU | On-demand, per GPU hour | Spot, per GPU hour |
|---|---|---|
| NVIDIA L40S 48GB | $0.55 | not listed |
| RTX 4090 24GB | $0.33 | $0.17 |
| RTX 6000 Ada 48GB | $0.77 | $0.39 |
| RTX 5090 32GB | $0.56 | not listed |
How Billing Works
- Model APIs: per token.
- Agent Sandbox and serverless GPU: per second.
- GPU instances: per hour per GPU.
The pages we read do not state a minimum top-up or monthly fee. Check the console for billing rules before you add funds.
Batch Discount And Savings
The pricing page says batch inference has an introductory 50% discount on input and output tokens for supported models. Spot GPUs cost about half of on-demand on the two models that list both. See the discount page.
Before You Pay
- Prices change, and the home page and pricing page showed different DeepSeek V4 Pro figures.
- Check taxes and any minimum charge at checkout.
- Prices for serverless GPU, bare metal and image or video APIs were not on the pricing page.
- Set a spending limit before large jobs.
How We Wrote This
This page is based on the official Novita AI website, pricing page and GPU page, read on 24 October 2026. We have not tested Novita AI ourselves, so we do not claim results. Speed, uptime and savings figures are the vendor's own claims. Prices, models and promotions change often, so confirm them on the official site before you pay. See how we review.