“Leading large language models in China are estimated to be about six months behind their U.S. counterparts in capability but are 10 times cheaper.”