Novita AI
Novita AI sells pay-per-token API access to open models such as gpt-oss-120b and DeepSeek V4 Pro, and rents GPU instances billed by the second.
Last updated
Novita AI runs an API for open models from makers such as OpenAI, Meta and DeepSeek, with separate prices per million input and output tokens. It also rents on-demand and spot GPU instances, timed to the second and settled every hour, and says it has 14 regions. Its terms say that, by default, it does not train its own models on your content.
| Fact | Value | Checked | Source |
|---|---|---|---|
| Billing for model APIs | Per token used, with input and output tokens priced separately | docs.novita.ai | |
| Billing for GPU instances | Timed to the second and settled every hour | docs.novita.ai | |
| Tax | Not included in the listed fees; the customer pays any tax that applies | novita.ai | |
| Training on your content | Not by default, according to the terms | novita.ai | |
| Data terms | Terms of service (section 10.2) and privacy policy | novita.ai |
Plans and prices
These are all the plans we list from Novita AI, each with the price we last checked on the provider's own page and the date of that check.
| Provider | Plan | Price | Billing | New-customer offer | Link |
|---|---|---|---|---|---|
| AI models | |||||
| Novita AI | OpenAI GPT OSS 120B OpenAI's gpt-oss-120b through Novita's API, priced per million tokens, with a 128K context window | $0.05 / $0.25 (about $0.10 per 1M tokens, 3 input : 1 output) per 1M tokens, input / output · Pay per use; input and output tokens priced separately checked 27 Oct 2026 | Per use | No offer | Visit |
| Novita AI | Llama 3.3 70B Instruct Meta's Llama 3.3 70B Instruct through Novita's API, priced per million tokens, with a 12K context window | $0.135 / $0.40 (about $0.20 per 1M tokens, 3 input : 1 output) per 1M tokens, input / output · Pay per use; input and output tokens priced separately checked 27 Oct 2026 | Per use | No offer | Visit |
| Novita AI | Qwen3.8 27B Qwen3.8 27B through Novita's API, priced per million tokens, with a 977K context window | $0.42 / $3 (about $1.07 per 1M tokens, 3 input : 1 output) per 1M tokens, input / output · Pay per use; input and output tokens priced separately checked 27 Oct 2026 | Per use | No offer | Visit |
| Novita AI | DeepSeek V4 Pro DeepSeek V4 Pro through Novita's API, priced per million tokens, with a 1M context window; cached input costs less | $1.60 / $3.20 (about $2.00 per 1M tokens, 3 input : 1 output) per 1M tokens, input / output · Pay per use; input and output tokens priced separately checked 27 Oct 2026 | Per use | No offer | Visit |
| GPU servers | |||||
| Novita AI | GPU instance, RTX 4090 24GB A GPU instance with one RTX 4090 (24 GB), priced per GPU by the hour and billed by the second; disks are billed separately. Fits 7–14B models | $0.33/hour (about $3.30 for 10 hours of use per month) (about $58 for 176 hours of use per month) (about $241/month) On demand, per GPU; billed by the second, storage extra checked 27 Oct 2026 | Per second | No offer | Visit |
| Novita AI | GPU instance, NVIDIA L40S 48GB A GPU instance with one NVIDIA L40S (48 GB), priced per GPU by the hour and billed by the second; disks are billed separately. | $0.55/hour (about $5.50 for 10 hours of use per month) (about $97 for 176 hours of use per month) (about $402/month) On demand, per GPU; billed by the second, storage extra checked 27 Oct 2026 | Per second | No offer | Visit |
| Novita AI | GPU instance, RTX 5090 32GB A GPU instance with one RTX 5090 (32 GB), priced per GPU by the hour and billed by the second; disks are billed separately. Fits 7–14B models | $0.56/hour (about $5.60 for 10 hours of use per month) (about $99 for 176 hours of use per month) (about $409/month) On demand, per GPU; billed by the second, storage extra checked 27 Oct 2026 | Per second | No offer | Visit |
| No plan matches these filters. Show all plans | |||||
Not shown: 0 plans whose data-centre locations we have not confirmed yet.
Monthly costs for hourly prices assume the server runs all month (730 hours), or less where a provider caps the monthly bill.
Data-centre locations
We have not confirmed where Novita AI runs its data centres yet.
Where we used it
Our guides that include this provider: