Managed models, priced per token. Idle time costs nothing.
Every model in the catalogue serves at one USD rate per million tokens, input and output the same. A quiet hour bills nothing. A busy hour bills the tokens it served.
from openai import OpenAI
client = OpenAI(
base_url="https://api.nucleus-ai.io/v1", # the one line that changes
api_key=NUCLEUS_API_KEY,
)
client.chat.completions.create(
model="nucleus://run_3Bz:train:0/sampler_weights/000240",
messages=[{"role": "user", "content": "Triage this ticket"}],
)One rate per model.
Input and output tokens cost the same, in USD per million.
A quiet hour bills nothing.
You pay for the tokens you serve. Idle time costs nothing.
Your fine-tune at the base rate.
A fine-tuned model serves at the same per-token rate as its base model, with no standing per-model fee.
Your SDK already works.
Swap the base URL and your OpenAI or Anthropic client keeps the rest of its code.
What a serverless endpoint costs
Two things decide the bill: the model's rate and the tokens you serve.
One rate per million tokens, by model group
Every base model carries a served-token price in USD per million, and input and output tokens cost the same. A fine-tuned model serves at its base model's rate. Media models meter the tokens of the encoded media: a megapixel image is roughly 4K tokens, a five-second 720p clip roughly 70K and a minute of audio roughly 3K.
Results from LLMs powered by Nucleus
sweep of the tracking feed
Shipment Delay Management / Logistics
from site reading to client report
Real-Time Field Data Collection / Field engineering
delays resolved without a call
Intelligent Fleet Optimisation / Fleet operations
Serverless, in four columns
Pricing
- USD per million tokens
- Input and output the same price
- Fine-tuned models at the base rate
- No standing per-model fee
- Idle time costs nothing
Models
- Language, 21
- Vision, 7
- Embedding and reranking, 6
- Image generation, 4
- Video generation, 3
- Audio, 8
- Media metered as tokens of the encoded media
Routes
- OpenAI-compatible completions and chat completions
- Anthropic-compatible messages
- Samplers for token-level sampling
- nucleus:// paths for your fine-tuned weights
- The infer scope on serving routes
Usage
- Served tokens reported for any date range
- Grouped by day, project or model
- Forecast before the bill and reconcile after it
Know what you pay
Serverless runs on the served-tokens meter. Everything on this page is priced from the same catalogue the pricing page reads.
The rest of Serve
Private Endpoints
OpenAI and Anthropic compatible
Dedicated GPUs
L40S to B200, by the second
Reserved capacity
Guaranteed terms, one to six months
Compatible APIs
DocsOne key, two SDK families
Read on
Serving and exporting
Take finished weights to production by serving them on Nucleus or exporting them to run anywhere.
Use the OpenAI SDK
Point any OpenAI client at Nucleus and chat with your fine-tuned model without changing your code.
Usage and audit
Meter token and storage consumption and read the organisation's audit trail.