Novita AI Pricing Calculator
Updated July 2026Pure usage (per-token inference + per-hour GPU + per-second sandbox), free to start
Novita AI pricing: Novita AI pricing 2026: per-token inference from $0.02/M, GPUs from $0.35/hr, bare-metal H100 $1.70/GPU/hr, and per-second agent sandboxes. Free to start. Use the free calculator below to enter your usage and get an instant all-in monthly estimate — including overages — and see which plan is cheapest for your needs.
How Novita AI prices: Pure usage (per-token inference + per-hour GPU + per-second sandbox), free to start.
How Novita AI prices
Pure usage (per-token inference + per-hour GPU + per-second sandbox), free to start
Your usage
💡 A busy chat product sends ~300M input tokens/mo (e.g. 3M requests × 100 tokens).
Per-1M input-token rate of the model you call. Pick the same model for output below.
💡 Output is usually a fraction of input — a mid-size app generates ~120M tokens/mo.
Per-1M output-token rate. Match the model you chose for the input rate.
💡 One always-on GPU ≈ 730 GPU-hr/mo; an 8-GPU bare-metal node full-time ≈ 5,840.
On-demand instance, spot, dedicated-endpoint, or bare-metal per-GPU-hour rate (same H100 is priced three ways).
💡 Text-to-Image (512×512, 5 steps) is $0.001/image; premium models like Hunyuan Image 3 reach $0.10.
Per-image rate for the generation/edit model you call.
💡 A 5s Hunyuan Video Fast clip is $0.30; a 5s Kling 2.5 Turbo clip is $0.35.
Per-finished-video rate for the model and duration you call (per-second models priced at a representative 5s clip).
💡 Fish Audio TTS is $15/1M characters; premium MiniMax HD voice is $100/1M.
Per-1M-character text-to-speech rate for the voice model you call.
💡 A 1-hr code job on 2 vCPU + 1 GiB RAM bills ~$0.0821 → ~$0.041/vCPU-hr at that config.
Monthly Estimate
Pay-as-you-go
$201.00
per month
Cost Breakdown
Annual
$2,412.00
Why Pay-as-you-go — $201.00/mo
- Pay-as-you-go at $201.00/mo.
- How we got there: 300 M tokens (LLM input tokens / mo) × 0 $/1M in (LLM input rate) = 81 $.
- How we got there: 120 M tokens (LLM output tokens / mo) × 1 $/1M out (LLM output rate) = 120 $.
- How we got there: 0 GPU-hr (GPU-hours / mo) × 2 $/GPU-hr (GPU type & packaging) = 0 $.
- How we got there: 0 images (Images / mo) × 0 $/image (Image model) = 0 $.
- How we got there: 0 videos (Videos / mo) × 0 $/video (Video model) = 0 $.
- How we got there: 0 M characters (Text-to-speech characters / mo) × 15 $/1M chars (TTS model) = 0 $.
- How we got there: 0 vCPU-hr (Agent-sandbox vCPU-hours / mo) × 0 (×) = 0 $.
- LLM output-token spend (120 $ × $1) is 60% of your bill — the lever that matters most.
- Need less? LLM input rate: Llama 3.1 8B — $0.02 drops you to Pay-as-you-go at $126.00/mo.
Budget range
Plan for usage swings, not just today's estimate.
Conservative
$160.50
150 M tokens/mo
Expected
$201.00
300 M tokens/mo
Aggressive
$282.00
600 M tokens/mo
Need a calculator like this on your pricing page?
Embed interactive pricing calculators on your website to help customers understand costs and boost conversions.
Get StartedAbout this Novita AI calculator
This calculator estimates your Novita AI cost from publicly available pricing. Actual costs may vary with your specific agreement, volume discounts, and usage patterns — always verify on the provider's official pricing page for the most current rates.
Novita AI pricing — frequently asked questions
How does Novita AI pricing work? ▼
Novita is pure pay-as-you-go. Model inference is billed per million tokens (LLMs), per image, per video-second, or per 1M characters (audio); GPUs are billed per hour (on-demand or spot); and agent sandboxes are billed per second on vCPU and memory. There is no monthly minimum and you can start for free.
How much does an NVIDIA H100 cost on Novita AI? ▼
It depends on the product. A dedicated inference endpoint on H100 80GB is $1.99 per GPU-hour, and an 8-GPU bare-metal H100 SXM node is $1.70 per GPU/hr. As of mid-2026 H100 is no longer listed on Novita's self-serve GPU-instance page, which carries RTX-class GPUs (RTX 4090 from $0.33/hr on-demand, $0.17/hr spot).
Does Novita AI have a free tier? ▼
Yes. Novita advertises 'Free to start, scales as you grow' — you can sign up and call the model APIs without a subscription, paying only for the tokens, images, GPU hours, or sandbox seconds you consume.
What is the cheapest LLM on Novita AI? ▼
Among flagship-class models, Llama 3.1 8B Instruct is among the cheapest at $0.02/M input and $0.05/M output. Larger models like DeepSeek V4 Pro run $1.6/M input and $3.2/M output. As of the 2026-07-06 capture, the base RTX 4090 24GB instance is $0.33/hr on-demand ($0.17/hr spot).
Is this Novita AI pricing calculator free and accurate? ▼
Yes — it's completely free with no signup. It uses the latest publicly available Novita AI pricing (verified July 2026). Actual costs may vary with volume discounts or enterprise terms — always confirm on Novita AI's official pricing page.