Meta aims for 55% score on Humanity's Last Exam by year-end

Meta: Meta Platforms develops and deploys large language models as part of its broader AI research efforts. Its Muse Spark series has emerged as a focal point for recent performance evaluations on challenging benchmarks. The company's model releases and benchmark results directly inform market sentiment around its ability to reach specific accuracy thresholds by year-end. Opus 5: Opus 5 is an Anthropic AI model focused on high-performance reasoning and problem-solving. It participates in head-to-head comparisons with other frontier systems on rigorous academic-style tests. Its results contribute to the broader context of Meta's year-end performance outlook. Claude Fable 5.1: Claude Fable 5.1 is an Anthropic large language model optimized for complex reasoning across multiple domains. It currently leads certain public evaluations of frontier AI capabilities. Its benchmark position serves as a reference point for assessing Meta's competitive standing in the news. Muse Spark series: The Muse Spark series comprises Meta's multimodal AI models designed for advanced reasoning tasks. Versions 1.1 and 1.3 have recently featured in public leaderboard comparisons on expert-crafted benchmarks. These releases drive trader attention to Meta's progress relative to competing frontier systems. Market Focus: Trader sentiment tracks upcoming model releases and minor benchmark protocol adjustments as potential swing factors in frontier AI progress. AI Benchmarking: Humanity's Last Exam evaluates closed-ended accuracy on expert-designed questions spanning mathematics, physics, and related fields where frontier models still face significant challenges. Competitive Landscape: Rapid iteration of multimodal and reasoning features determines whether models can meet or exceed key performance thresholds before the end of the year.