now this is the kind of model benchmark i care about
How well can models design icons? @andpoul and I benchmarked frontier and open weight models
The results: they’re steadily improving execution, their choices often overly literal
顯示更多
