Meta’s Muse Spark series, particularly the 1.1 variant, has posted scores near 62% on Humanity’s Last Exam leaderboards as of early October 2026, placing it ahead of several GPT-5 and Gemini variants but behind Anthropic’s leading Claude Opus 5.5 and Fable 5.1 models at 65%. This performance stems from Meta’s focus on efficient training and tool-augmented reasoning for its large language models, though the benchmark’s expert-authored questions across math, sciences, and humanities continue to highlight gaps versus closed frontier systems. Traders weigh Meta’s rapid iteration cadence and open-weight strategy against the risk that remaining 2026 releases may not close the gap to 65%+ before year-end resolution on the official HLE leaderboard.