▶Meter is a research nonprofit focused on measuring catastrophic risks from AI autonomy, serving as an early warning system for when AI can improve itself.Apr 2026
▶Meter has become an influential, industry-standard benchmark for measuring the performance of frontier AI models, with its 'time horizon' chart being widely cited.Apr 2026
▶Meter's benchmarks specifically focus on complex tasks typical for an engineer at a frontier AI lab, such as software engineering and fine-tuning models.Apr 2026
▶Major AI labs are generally cooperative, providing Meter with the necessary access to models for its research.Apr 2026
▶There is a contrast between Meter's findings of exponential growth in frontier AI capabilities and a separate Meter report cited as showing negligible or even negative productivity impact from current AI coding tools in enterprises.Mar 2026
▶The measured rate of AI progress is accelerating rapidly; a late 2023 Meter paper noted a seven-month doubling time for AI coherence, while more recent claims citing Meter state the doubling time for capabilities is now approximately four months.Mar–Apr 2026
▶Meter's benchmarks face a methodological tension: while they track a steep upward trend in AI agent task length, they are also reportedly becoming 'saturated' because models are improving so quickly that it is difficult to create sufficiently challenging evaluation tasks.Apr 2026
Sign up free to see the full intelligence report
Get started free