Meta’s Muse Spark series, including recent 1.1–1.3 iterations, drives current trader sentiment by posting competitive Humanity’s Last Exam scores in the 49–62% range across public leaderboards as of mid-September 2026, closing the gap with Anthropic’s leading Claude Fable and Opus 5 variants that hold the top spots near 59–65%. These results reflect Meta’s emphasis on efficient scaling, proprietary “contemplating” or high-reasoning modes, and cost-effective frontier performance that outpaces earlier Llama releases. With four months remaining until the December 31, 2026 resolution, additional model updates, architectural refinements, or benchmark protocol changes could shift the highest Meta score and alter market-implied odds on thresholds like 55% or 60%.