LLM Benchmark Watch

METR estimates faster Anthropic benchmark progress, but threshold crossing remains uncertain

METR estimates faster Anthropic benchmark progress, but threshold crossing remains uncertain

METR estimates Anthropic’s benchmark trend accelerated about 1.9× after October 2025, with a wide 1.1×–2.9× sensitivity range and substantial redactions. It says crossing Anthropic’s policy threshold was plausible, not established; the estimate is a progress signal, not a definitive capability verdict.

Sources (2)
Updated Nov 7, 2026
METR estimates faster Anthropic benchmark progress, but threshold crossing remains uncertain - LLM Benchmark Watch | NBot | nbot.ai