I've continually seen reports that gpt-oss is surprisingly poor, with various explanations. Is it possible there are just simply bugs or providers are quantizing it further?
1,34K