A little context goes a long way.

What is Nemotron API access?

An API lets your software send a prompt to a model and receive a response. Rittenhouse Compute is preparing access to a small selection of NVIDIA Nemotron models through a shared interface. Our intended gateway is New API, connected to OpenRouter.

The pilot is being prepared. There is no public inference endpoint or API key signup on this site yet. Compare the candidate models.

Technical references: New API and OpenRouter’s API documentation.

What goes into a token estimate?

Input is the text you send: instructions, source material, and conversation context. Output is the text the model generates. Token counts depend on the model’s tokenizer; a token is not necessarily a word.

At the draft Super rates, 2,000 input tokens and 500 output tokens total $0.001125. A thousand identical requests would total $1.125, before any separately agreed integration work. These are estimates, not approved commercial terms.

What makes an extraction inspectable?

A useful extracted field keeps its source. For example, “observational” should link to the sentence that identifies the study design, rather than appear as an unexplained label. Missing or ambiguous information should remain visible.

Our prepared protocol example demonstrates that interaction with synthetic text. It does not measure model accuracy. A real evaluation would compare extracted fields against reviewed reference values and report both successes and errors.

What should you bring to a pilot conversation?

• A specific task: what goes in, and what a useful result looks like.
• A small public or synthetic sample that can be reviewed safely.
• An estimate of request volume and a budget range.
• The failure cases that matter and who can review the results.

Do not submit patient records, confidential protocols, or other sensitive data through the inquiry form. Begin with a non-sensitive description.