Institutions / METR

METR

Model Evaluation & Threat Research

metr.org · @METR_Evals

Their benchmarks in the atlas

METR Task-Completion Time Horizonsby METRTask level · 50% task-completion time horizon
17 hClaude Mythos Preview (early) · 7 Apr 2026
Unrated
RE-Benchby METRTask level · Human-expert percentile matched (8-hour budget)
37%Claude 3.5 Sonnet (New) (Modular scaffold) · 31 Jan 2025
Emerging

Work their benchmarks cover

Software engineering ML and research engineering Security

Embed this card

The image names the institutions behind the numbers and links back to this page.

HTML:

<a href="https://canagentswork.com/institutions/metr/"><img src="https://canagentswork.com/og/institutions-metr.png" width="600" height="315" alt="METR's agent benchmarks, mapped onto the world of work:" loading="lazy"></a>

Markdown:

[![METR's agent benchmarks, mapped onto the world of work:](https://canagentswork.com/og/institutions-metr.png)](https://canagentswork.com/institutions/metr/)

For METR: link to this page

You can link to this page from your site. Copy this HTML:

<a href="https://canagentswork.com/institutions/metr/">Featured in the Work Atlas — Can Agents Work?</a>

If a summary or number on this page is wrong, tell us. We correct errors within 48 hours of a report. To grant permission to show your logo, contact us; until then we show only your name.