Gemma ran at 50/tops
Qwen 27B? was way slower , 5/tops
8B models run perfectly fine, but are mostly useless for chat and agents.
8B is only good for specialists. Like one 8B model that can only write and correct python3 code.
And then only in English.
Gemma ran at 50/tops Qwen 27B? was way slower , 5/tops 8B models run perfectly fine, but are mostly useless for chat and agents. 8B is only good for specialists. Like one 8B model that can only write and correct python3 code. And then only in English.