Meta: Muse Wonder

Meta's Muse Wonder, a frontier reasoning model created on October 12, 2026, demonstrates a strong performance profile, particularly in its reliability and accuracy across complex tasks. The model exhibits competitive response times, ranking in the 45th percentile for speed across four benchmarks. However, it is positioned at premium pricing levels, falling into the 5th percentile for cost. A standout feature is its exceptional reliability, achieving a 100% success rate across all benchmarks, indicating minimal technical failures and consistent evaluable responses. In terms of specific capabilities, Muse Wonder excels in Instruction Following, achieving 89.0% accuracy (94th percentile), and Coding, with 95.0% accuracy (83rd percentile). Its Reasoning and Mathematics capabilities are also robust, scoring 92.0% (64th percentile) and 94.0% (69th percentile) respectively. While its accuracy is consistently high, the model's duration for these tasks, particularly Coding and Mathematics, is on the higher side, contributing to its average speed ranking. Its primary strength lies in its high accuracy and unwavering reliability across diverse and complex agentic tasks, making it a dependable choice despite its premium cost and moderate speed.