World Model · podcast knowledge graph

METR company

16 mentions · across 9 shows · also seen as concept

Model Evaluation and Threat Research (METR), is a nonprofit research institute, based in Berkeley, California, that evaluates frontier AI models' capabilities to carry out long-horizon, agentic tasks that some researchers argue could pose catastrophic risks to society. METR has worked with leading AI companies to conduct pre-deployment model evaluations and contribute to system cards, including OpenAI's o3, o4-mini, GPT-4o and GPT-4.5, and Anthropic's Claude models. Following the 2026 OpenAI agent cyberattacks, METR was invited to conduct an independent investigation of the incident. On August 26, 2026 it published a report of agents’ behavior, reasoning and collaboration during the incident

https://en.wikipedia.org/wiki/METR

Relationship map

Connections 13

mentioned 10

works at 3

Heard in

Mentioned in episodes

All extracted evidence for this entity (13)