Are those with «thinking» or without?

Sonnet 3.7's 70% is without thinking, see https://www.anthropic.com/news/claude-3-7-sonnet

The thinking tokens (even just 1024) make a massive difference in real world tasks with 3.7 in my experience

based on their release cadence, I suspect that o4-mini will compete on price, performance, and context length with the rest of these models.

With