3 ms·

I haven't experienced this personally, but I have stuck with pretty mainstream models like llama, gemma, deepseek, etc.