Back
Credibility Rating
4/5
High(4)High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: METR
METR is a leading third-party AI safety evaluation organization whose work on autonomous capability benchmarks and catastrophic risk assessments directly informs AI lab safety policies and government AI governance frameworks.
Metadata
Importance: 82/100homepage
Summary
METR is an organization conducting research and evaluations to assess the capabilities and risks of frontier AI systems, focusing on autonomous task completion, AI self-improvement risks, and evaluation integrity. They have developed the 'Time Horizon' metric measuring how long AI agents can autonomously complete software tasks, showing exponential growth over recent years. They work with major AI labs including OpenAI, Anthropic, and Amazon to evaluate catastrophic risk potential.
Key Points
- •Developed the 'Task-Completion Time Horizon' metric showing exponential increase in AI autonomous task completion capability over 6 years.
- •Conducts third-party evaluations for frontier AI labs assessing risks from AI self-improvement, rogue replication, and sabotage of oversight.
- •Empirical finding that AI tools can make experienced open-source developers 19% slower, challenging assumptions about productivity gains.
- •Focuses on evaluating broad autonomous capabilities and AI's ability to accelerate AI R&D—key precursors to loss-of-control scenarios.
- •Studies potential AI behaviors that threaten evaluation integrity and develops mitigations for such behaviors.
Cited by 33 pages
Cached Content Preview
HTTP 200Fetched Feb 27, 202616 KB
[](https://metr.org/)
- [Research](https://metr.org/research)
- [Notes](https://metr.org/notes)
- [Updates](https://metr.org/blog)
- [About](https://metr.org/about)
- [Donate](https://metr.org/donate)
- [Careers](https://metr.org/careers)
Menu
Model Evaluation & Threat Research
METR conducts research and evaluations to improve public understanding of the capabilities and risks of frontier AI systems.
[Our research](https://metr.org/research) [Careers](https://metr.org/careers)
We’ve worked with
[202020212022202320242025030 min1 hour2 hours3 hours4 hours5 hours6 hours7 hours8 hours9 hours10 hours11 hours12 hours13 hours14 hours15 hoursFix complex bug in ML research codebaseExploit a vulnerable Ethereum smart contractTrain adversarially robust image modelExploit a buffer overflow in libiec61850Fix bugs in small Python librariesGPT-2GPT-3GPT-3.5GPT-4GPT-4 Nov '23Claude 3 OpusGPT-4 TurboGPT-4oClaude 3.5 Sonnet (Old)o1-previewClaude 3.5 Sonnet (New)o1Claude 3.7 Sonneto3Claude Opus 4Claude Opus 4.1GPT-5Gemini 3 ProGPT-5.1-Codex-MaxClaude Opus 4.5GPT-5.2 (high)ClaudeOpus 4.6GPT-5.3-Codex (high)](https://metr.org/time-horizons/)
Time Horizon 1.1 (Current)TH 1.1
Time Horizon 1.1 (Current)
Follows the same methodology described in the initial paper, but with a larger task suite. See [release announcement.](https://metr.org/blog/2026-1-29-time-horizon-1-1/)
Time Horizon 1.0 (Mar 2025)
Original time horizon computations. Calculated for models from 2019 through Nov 2025, following the methods described in the original time horizon paper.
Log ScaleLinear Scale
50% Success80% Success
[**Task-Completion Time Horizons of Frontier AI Models** \\
We propose measuring AI performance in terms of the length of software tasks AI agents can complete. We show an exponential increase in this time horizon metric over the past 6 years.](https://metr.org/time-horizons/)
[Read paper](https://arxiv.org/abs/2503.14499) [View repo ](https://github.com/METR/eval-analysis-public)
## Featured research
Our AI evaluations research focuses on assessing broad autonomous capabilities and the ability of AI systems to accelerate AI R&D. We also study potential AI behavior that threatens the integrity of evaluations and mitigations for such behavior.
[View all research](https://metr.org/research)
- [General](https://metr.org/#research-for-the-public)
- [Technical](https://metr.org/#research-for-researchers)
- [Policy](https://metr.org/#research-for-policy-researchers)
- [View all research](https://metr.org/research)
### GPT-5.1 Evaluation Results
We evalua
... (truncated, 16 KB total)Resource ID:
45370a5153534152 | Stable ID: NWFiOTg0MD