Lumen
A ready-to-use AI service, hosted by NCSA: a curated set of models you can use anytime — from the web chat or straight from your own code — behind a single Illinois login.
What it is
Lumen is a managed AI service that NCSA runs for you, so there is nothing to install or operate. It puts a curated set of models behind one Illinois login and makes them available two ways at once: a full web chat interface for everyday use, and an OpenAI-compatible API you can call from your own scripts and applications.
Because access, usage limits, metering, and reporting are handled centrally, you — and your programs — can start using models right away and switch freely between them, without provisioning any infrastructure of your own.
When to use it
Good fit
- You want ready-to-use models without running any infrastructure yourself.
- You want the same models from a chat UI and from your own code.
- You want to sign in with your Illinois NetID (CILogon).
- You want usage visibility and optional per-user or per-group usage limits.
Look elsewhere if
- You need commercial models (GPT, Claude, Gemini) — see IllinoisChat.
- You want document Q&A with citations — see IllinoisChat.
- You need large offline batch jobs — see LLMFlux.
- You just want to deploy an open model on HPC — see LLM Hub.
Key capabilities
| Models | A curated set of open-weight models, all hosted on Illinois/NCSA. Choose whichever model fits your task — per conversation in the chat, or per request from your code — with no setup on your part. |
|---|---|
| Signing in | Sign in with your Illinois NetID (single sign-on); for scripts and apps, generate a personal API key. You’re placed in the right group automatically based on your campus affiliation, so the appropriate models and limits apply. |
| Your usage | Check your usage any time on a personal page — your remaining balance, the models you’ve used, and your API keys. Group administrators get aggregate dashboards showing usage across their team. |
| Cost | No charge to use. Like other Illinois Computes resources, Lumen is provided to campus researchers at no cost — no money changes hands. Usage is metered in coins, a virtual allowance rather than currency: each model has a coin rate per 1M input/output tokens (and per audio-hour for speech), so admins can set fair per-user or per-group quotas. |
Data & hosting note
Lumen serves only models hosted on Illinois/NCSA infrastructure — nothing is routed to an outside vendor — so your prompts and data stay within University control. That makes it a good choice for sensitive work, though you should still follow your unit’s data-classification guidance for regulated data (FERPA, HIPAA, export-controlled, and the like).
What’s recorded: only your web-chat conversations are saved — so you have a chat history — and you can delete them at any time. Anything you send through the API is not recorded: Lumen keeps only usage totals (tokens, cost, timing), never your prompts or the model’s replies.