Model Evaluation & Threat Research
We help AI companies and wider society understand the capabilities and risks of advanced AI systems.
Our research Careers
Frontier Risk Report
(February to March 2026)
A pilot assessment of rogue deployment risk at frontier AI companies.
Non-public information and model access provided by:
Anthropic Google DeepMind Meta OpenAI
Read report
Time Horizon 1.1 (Current)
Follows the same methodology described in the initial paper, but with a larger task suite. See release announcement.
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.

Task-Completion Time Horizons of Frontier AI Models

We propose measuring AI performance in terms of the length of tasks AI agents can complete. We show an exponential increase in this time horizon metric over the past 6 years.
Read paper View repo

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
Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity

Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity

A survey of 349 technical workers finds a median 1.4–2x self-reported change in value of work due to AI tools, expected to grow over time, though there are reasons to be skeptical of the magnitude.

Read more
Early Work on Monitorability Evaluations

Early Work on Monitorability Evaluations

We show preliminary results on a prototype evaluation that tests monitors' ability to catch AI agents doing side tasks, and AI agents' ability to bypass this monitoring.

Read more
How Does Time Horizon Vary Across Domains?

How Does Time Horizon Vary Across Domains?

We build on our time-horizon work and analyze 9 benchmarks for scientific reasoning, math, robotics, computer use, and self-driving in terms of time-horizon trends; we observe generally similar rates of improvement to the 7-month doubling time in our original time-horizon work.

Read more
MALT

MALT

A dataset of natural and prompted examples of behaviors that threaten evaluation integrity, like generalized reward hacking or sandbagging

Read more
Hawk

Hawk

Our open-source platform for running AI agent evaluations at scale, built upon Inspect AI

View website

Measuring autonomous AI capabilities — resource collection

An index of our research and guidance on how to measure AI systems' ability to autonomously complete a wide range of multi-hour tasks

Read more

Common Elements of Frontier AI Safety Policies

An analysis of the shared components across twelve published frontier AI safety policies, including capability thresholds, model weight security, and deployment mitigations

Read more PDF

Frontier AI Safety Policies

A list of AI companies' frontier safety policies intended to evaluate and manage severe AI risks

Read more

What should companies share about risks from frontier AI models?

We describe areas for risk transparency and specific technical questions that a frontier AI developer could answer.

Read more

Partnerships

We have previously partnered with OpenAI, Anthropic, Google DeepMind, Meta and Amazon to pilot frontier risk assessments. These companies have also provided access and tokens, which we use for evaluations, research, and engineering.

We are also part of the NIST AI Safety Institute Consortium and California Cybersecurity Task Force, are partnering with the AI Security Institute, and provide technical assistance to the European AI Office.

Risk Assessment

Our work assessing risks from frontier AI systems — including the Frontier Risk Report, independent reviews of AI developers' risk assessments, and capability evaluations of frontier models.

View all evaluation reports

METR does not accept compensation for this work.

Companies such as OpenAI, Anthropic, and xAI have provided access and tokens, which we use for evaluations, research, and engineering. We also occasionally evaluate models independently after they are released, without involvement from the model's developers. Recent public reports resulting from this work are above, with additional discussion in the respective system cards.

Frontier AI Safety Policies

We advise AI developers and governments on implementing risk assessment methodologies for AI. For example, we have advised developers on Frontier AI Safety Policies.

Resources on FSPs
Deepmind logo
Google
OpenAI logo
OpenAI
Anthropic logo
Anthropic

METR in the press

View all press coverage

Recent

Metrics of Agent Ability
July 24, 2026

Tom Cunningham surveys metrics for comparing AI agent capability as performance changes with expenditure and relative to human performance.

Read more
The Economics of Recursive Self-Improvement
July 22, 2026

Parker Whitfill and Tom Cunningham highlight context and takeaways from a new paper modeling how AI may accelerate AI R&D, and whether feedback effects could be strong enough for self-sustaining acceleration.

Read more
Expenditure Horizon: Measuring Optimization Ability, with an Application to NanoGPT
July 21, 2026

We propose a measure of an AI agent’s optimization ability with an "expenditure horizon." We give an empirical illustration from the NanoGPT speedrun.

Read more
Expenditure Horizon: Measuring Optimization Ability, with an Application to NanoGPT
Because 8 ≈ e², Anthropic's researcher uplift is plausibly >2x
July 8, 2026

Anthropic reported that contributors merge 8x as much code per day as before AI. Thomas Kwa argues that under standard economic modeling assumptions, this implies the uplift of individual Anthropic researchers from coding agents alone is above 2x.

Read more
Because 8 ≈ e², Anthropic's researcher uplift is plausibly >2x
Summary of METR's predeployment evaluation of GPT-5.6 Sol
June 26, 2026

A summary of METR's independent, predeployment evaluation of GPT-5.6 Sol

Read more
Frontier Risk Report (February to March 2026)
May 19, 2026

A pilot assessment of rogue deployment risk at frontier AI companies. Starting in February 2026, METR conducted a pilot exercise to assess misalignment risks from AI agents used inside frontier AI developers, with participation from Anthropic, Google, Meta, and OpenAI.

Read more
Frontier Risk Report (February to March 2026)