# Intelligence per Watt leaderboard

external benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelscale.dev/leaderboards/benchmarks/external-intelligence-per-watt

Intelligence per Watt (local AI energy efficiency). Every model the catalog carries a published Intelligence per Watt value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | external |
| Measure | Intelligence per joule (higher is better) |
| Tasks | Local LLM query efficiency benchmark suite |
| Difficulty | Energy-efficient local inference |
| Published by | [Intelligence per Watt: Measuring Intelligence Efficiency of Local AI](https://arxiv.org/abs/2511.07885) |

A local-AI efficiency metric from Stanford Hazy Research Lab and Scaling Intelligence Lab measuring task accuracy delivered per joule of energy consumed per query on local hardware.

## Intelligence per Watt ranking

1 model with a published Intelligence per Watt value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.

| Rank | Model | Provider | Intelligence per joule (higher is better) | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [Inkling 2 Mini](https://modelscale.dev/models/inkling-2-mini) | Thinking Machines Lab | 0.703 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
