Sustainability & AI
Frugal AI
Designing AI systems to achieve acceptable performance with minimal data, compute and energy, countering ever-larger models.
Definition
An approach to AI that aims to deliver needed functionality with minimal data, computation and energy — questioning at the outset whether a large model, or AI at all, is justified.
Quick reference
At a glance
- Subject
- Sustainability & AI
- Editorial status
- Editorial draft
- Definition status
- Emerging
- Last updated
- 21 August 2026
References
This source provides part of the technical or institutional basis for the definition.
This source supports the explanation of how the term is applied, measured or governed in practice.
Overview
What it means
Frugal AI extends sufficiency thinking to AI: right-size the model, reuse pre-trained components, prefer simple methods where they work, and weigh marginal accuracy gains against marginal resource costs. It aligns with Green AI's efficiency agenda and with life-cycle assessment of AI systems.
How it is used
Practices include benchmarking small against large models before scaling, deploying distilled or quantised models, using retrieval over regeneration, and setting compute budgets for projects. It is promoted in research policy circles as a counterweight to scale-driven development.
Why it matters
Frugal AI is the demand-side of sustainable AI: instead of only making computation cleaner, it asks how much computation a task actually needs. In a resource-constrained transition, that question is itself a sustainability intervention.
Review
Help keep this definition useful and accurate.
Submitted reviews stay private until an editor decides whether to accept and attribute them.
Endorse this definition
Confirm what works
Endorse the current wording when it is accurate and useful. Suggested changes use the separate editorial form below.