Claude 3.7 Sonnet

Anthropic's first hybrid reasoning model, letting users toggle standard and extended-thinking modes. Anthropic publishes no single-attempt extended-thinking GPQA figure for this model, so the standard-mode score is recorded.