Our Smallest, Most Efficient Model: Muse Ember

Muse Ember, a new compact AI model, has been released as the most cost-efficient option for complex reasoning tasks like tool use, coding, and document parsing, delivering better intelligence-per-dollar performance than Claude Haiku 5.5 and GPT-6 Luna. The model achieves comparable reasoning scores to Haiku at approximately one-third the cost and outperforms all Luna settings across various benchmarks including agent automation, coding tasks, and multimodal work.