Definition
The accumulation of unnecessary size, context, or compute cost in an AI model or the systems built around it — without a proportional gain in real-world usefulness.
Symptoms: rising inference cost, slower responses, inconsistent quality across sessions, growing operational overhead that outpaces the value it produces.
Field notes
- Users of major AI products have reported the same output feeling noticeably worse over time, even as the underlying systems keep growing. — reported spring 2026
- Flat-rate AI plans across several large providers were replaced with usage-based billing within weeks of each other, because unconstrained model usage stopped being financially sustainable. — reported spring 2026
- Analysts tracking data-center electricity use have started describing the growth curve in blunt terms — waste, excess — without yet having one word for the pattern. — reported summer 2026
Used in a sentence
Our inference bill tripled but the eval scores didn't move — classic model bloat.
That update wasn't a feature. It was model bloat with a changelog.
We need a model-bloat audit before the next training run.