Comment by joshrw

Chinese models optimize for benchmarks and do poorly in real-world tasks