Muse Spark 1.1: Meta gains 8 Intelligence Index points in three months (artificialanalysis.ai)

Meta's Muse Spark 1.1 has achieved a notable advancement, increasing its score on the Artificial Analysis Intelligence Index from 43 to 51 within just three months. This performance gain, driven by improvements in Scientific Reasoning, Coding, and Knowledge, positions Muse Spark 1.1 competitively alongside other leading models like GLM-5.2 and GPT-5.4. Key enhancements include a significant rise in agentic knowledge work, marking substantial progress in coding tasks, where it ranks third overall. Significantly, Muse Spark 1.1 not only excels in terms of intelligence but also stands out for its cost efficiency, using only 94 million output tokens for generating its Intelligence Index score, which translates to an estimated task cost of $0.26 – the lowest among its competitors at this performance level. Despite a slight dip in accuracy, its hallucination rate dropped markedly, demonstrating an effective approach to reducing unreliable outputs. With an impressive context window of 1 million tokens and strong output speed, Muse Spark 1.1 reflects Meta's commitment to advancing AI capabilities while offering economical solutions for developers and researchers in the AI/ML community.