METR and OpenAI state that artificial intelligence models are accelerating progress in AI research. METR estimates that Anthropic's benchmark testing pace has nearly doubled, while OpenAI reports improvements in task success rates without human intervention, though performance varies significantly across different tasks.