Anthropic has been very secretive about the supposed synthetic data they used to train 3.5 Sonnet. Given how good the model is terms of the quality vs speed tr

Anthropic has been very secretive about the supposed synthetic data they used to train 3.5 Sonnet. Given how good the model is terms of the quality vs speed tradeoff, they must have something.