Managed open-model API
Novita AI Endpoint Pricing Context
Compare public token offers before quoting an exact Novita route; unresolved, quote-only and dedicated models are excluded from this token shortlist.
Open-model prices to compare
Start with the public token rates below, then quote the exact Novita model and region you plan to use. Lowest list price is not always lowest production cost once latency, retries, context limits, and support matter.
| Model | Provider | Input | Output |
|---|---|---|---|
| DeepSeek-OCR 2 | novita | $0.03 | $0.03 |
| Llama 3.1 8B Instruct | novita | $0.02 | $0.05 |
| Llama 3 8B Stheno V3.2 | novita | $0.05 | $0.05 |
| AutoGLM-Phone-9B-Multilingual | novita | $0.035 | $0.138 |
| Qwen3.5 4B | castform | $0.03 | $0.15 |
| Llama 3 8B Instruct Lite | together | $0.14 | $0.14 |
| Qwen3.5 9B | together | $0.17 | $0.25 |
| Llama 4 Scout 17B 16E Instruct | groq | $0.11 | $0.34 |
| GLM-4.7-Flash | novita | $0.07 | $0.40 |
| Llama 3.3 70B Instruct | novita | $0.135 | $0.40 |
| Qwen2.5 7B Instruct Turbo | together | $0.30 | $0.30 |
| Qwen3.8 Flash | novita | $0.15 | $0.47 |
When Novita belongs in the quote set
Open-model experimentation
Use one managed API to test Llama, Qwen, DeepSeek-style, image, and video models before committing to a host.
Operational simplicity
A single endpoint can be cheaper than engineering time spent maintaining several vendor integrations.
Burst traffic
Managed APIs are easier when usage is spiky and you cannot keep rented GPUs busy.
Fallback routing
Treat it as another route in a provider failover plan, then measure latency and quality for your workload.