Meta’s frontier models, including the Muse Spark and Muse Wonder series, have posted competitive but not leading results on Humanity’s Last Exam (HLE), a 2,500-question expert-level benchmark spanning over 100 academic subjects and designed to resist simple retrieval. As of mid-October 2026, Muse Spark 1.1 reached 62.1% on one major leaderboard, trailing Anthropic’s Claude Fable 5.1 and Opus 5.5 variants that top out near 65%. Rapid capability gains across the industry—driven by larger models, improved reasoning techniques, and tool use—have lifted top HLE scores by roughly 30 points in the past year. Meta’s next model releases, developer conferences, or internal scaling updates could narrow the gap, while any saturation or variant-specific scoring rules will shape the final 2026 outcome. Traders should monitor official Meta announcements and independent benchmark updates for shifts in relative positioning.