AI researcher Mehu Mohan found Meta's Muse Wonderful model's performance to be inferior to that of Anthropic's Opus 5.5