Gpt-oss comes only in 4.5 bit quant. This is the native model, so there's no fp16 original