METR
AI research organization that evaluates and benchmarks frontier AI systems for autonomous task performance
metr.org ↗📍 Berkeley, California, USA
Verified Data
“Estimated 30k-100k visitors (niche research site, categorized under Philosophy/Science on SimilarWeb).”
“research task horizons reported to be doubling every 7 months.”
Company Profile
Contact
Strategic Analysis
Strategy
Independent non-profit model maintaining neutrality by not charging fees for evaluations, funded through grants and donations. Focus on being the trusted third-party evaluator for frontier AI models across major developers and government agencies. Spun off from Alignment Research Center to establish independence and scale evaluation capabilities.
Tactics
Direct partnerships with major AI labs (OpenAI, Anthropic, Google DeepMind) for model evaluations. Technical assistance contracts with government bodies like the European AI Office. Open research publication and dataset releases (like MALT) to establish thought leadership. Grant funding strategy through prestigious programs like The Audacious Project.
Competitive Positioning
Positioned as the independent, neutral evaluator in AI safety space, differentiating from internal company evaluations or academic research with limited scale. Competes with other AI safety organizations but maintains unique position through non-profit status and direct access to frontier models from major labs.
Marketing Approach
Research-led approach through publishing studies and datasets that generate media coverage. Direct engagement with policymakers and AI developers rather than broad consumer marketing. Thought leadership through controversial findings (like AI coding productivity study) that drive industry discussion.
Notable
Spun off from Alignment Research Center (ARC), funded through The Audacious Project
🔗 Source ↗Tech Stack
Python
PyTorch
React
Netlify
GitHubRecent News
Related AI Research Companies
Discovery Sources
Signals
“Estimated 30k-100k visitors (niche research site, categorized under Philosophy/Science on SimilarWeb).”
“research task horizons reported to be doubling every 7 months.”
“~100,000 - 150,000 monthly visits (Ranked #30 in Science and Education > Philosophy category)”
“Estimated $38M+ committed via The Audacious Project (TED) and additional grants from Open Philanthropy/Coefficient Giving. Stage: Non-profit / Grant-funded.”
Evidence
We find that roughly half of test-passing SWE-bench Verified PRs... would not be merged into main by repo maintainers.
LLM performance is much worse under the stringent success criterion.
4+ major AI lab partners plus multiple government agencies
47 employees
Rebranded from ARC Evals in 2024, expanded to 47 staff, established regulatory partnerships
$38M via The Audacious Project
38 employees