METR
Autonomy/task-horizon evals; 'time-horizon doubling'
METR (Model Evaluation & Threat Research) is a leading independent evaluator of dangerous autonomous capabilities, known for its 2025 finding that the length of tasks AI can complete autonomously is doubling roughly every ~7 months. Labs including OpenAI and Anthropic grant it pre-release access. Its task-horizon metric became a widely cited progress indicator.
Signature result
Task-horizon ~7mo doubling
Role
Autonomy red-teaming
How it fits the stack
METR with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.
METR in the AI stack. METR with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.
Graph data (text) — 2 entities, 1 relationships
- METR —partners with→ Anthropic
Context — capital, rivals, policy · · 1