Meta’s Muse Spark series, particularly the 1.1 and 1.3 variants, has driven recent trader sentiment by posting HLE scores in the 49–62% range depending on evaluation protocols such as tool use or high-reasoning modes. These results place Meta within a few points of Anthropic’s leading Claude Fable 5.1 (around 59–65%) while surpassing several OpenAI GPT-5 configurations on certain leaderboards. Momentum stems from Meta’s scaled training infrastructure, iterative post-training on expert-level data, and adaptive reasoning techniques that have narrowed the gap faster than earlier Llama releases. With three months remaining until the December 31, 2026 resolution, key catalysts include any fall model drops, new HLE versions with held-out questions, and disclosed scaling gains that could push Meta’s peak result above 60% or higher.