Unlimited,
explained honestly.
"Unlimited" is a word that's been abused. Here's exactly what it means at Denizen Blu — the promise, the one boundary, and why it exists.
Unlimited interactive use, by one person
A seat is one human. That human can work all day, every day, across every model up to 70B, and never see a token counter, an overage, or a wall. We designed the whole system around that experience.
Why there's a benchmark at all
Real infrastructure has real limits. If a single seat could drive unbounded automated traffic, one account could degrade service for everyone else — and the honest thing is to say so up front rather than hide it in a rate-limit error at 2am.
So we publish a benchmark: 50 million output tokens per month, per seat. This is deliberately generous — it's far more than a person can produce through interactive work in a month. It exists as a reference point for the rare outlier, not as a budget you should watch.
If a seat runs substantially above the benchmark, here is the entire consequence: that seat may receive reduced scheduling priority during peak demand. Requests still complete. Nothing is billed. Nothing is cut off. As demand eases, priority returns to normal.
What Fair Use never does
Charge you per token. Impose a hard cap. Terminate a session. Bill you for going over. Surprise you.
What it can do, rarely
Lower scheduling priority for a single extreme-outlier seat during peak load — temporarily, and never as a bill.
One seat is one person
Unmetered pricing only works when a seat maps to a human. A few common questions about where the line sits.
No — a seat is one person. Sharing a single seat as a shared firehose for many people (or many bots) is the one thing the model isn't built for. Add seats for teammates; it's $1.99/month (or $11/year) each and takes seconds.
Yes, commercial and production use is included. For sustained high-concurrency production traffic, enterprise plans provide dedicated capacity and higher concurrency so you're never competing with your own interactive work.
A person driving the model — through the playground, your app, an IDE, or an agent they're actively working with. It's the opposite of an unattended script hammering the API around the clock from a single shared key.
If we adjust it, we'll publish the change and the reasoning here first. Transparency is the point of having a benchmark at all — moving it quietly would defeat that.
No meter. No fine print.
That's the whole policy. If any of it ever feels less than honest, tell us — we'll fix the policy or fix the wording.