Most of those inference providers are American, and China is actually at a disadvantage here because of export restrictions - US companies are using newer and more efficient chips.
If it’s newer and efficient then why is the api more expensive?
If it’s newer and efficient then why is the api more expensive?