Realistically it can't match Fable, let alone Opus today. It can however beat basically all of the models from 1 year ago. If last years state of the art is "good enough" for you today then local is great.
What open-weight models can match Opus 4.x or GPT 5.4 performance at a reasonable throughput?
What open-weight models can match Opus 4.x or GPT 5.4 performance at a reasonable throughput?