Meta's internal codename "Watermelon" mapped to the Muse Wonder family released October 12, 2026, marking the company's first confirmed frontier large language model with closed API access and tiered subscriptions. The launch followed months of internal updates from Alexandr Wang highlighting roughly 10× compute scaling over Muse Spark, benchmark parity claims with GPT-5.5-class systems, and emphasis on agentic capabilities like active memory and multi-step professional tasks. Within a day, Muse Wonder (Max) entered the Text Arena leaderboard at #2 with scores reported between 1513 and 1539, a sharp gain from prior Muse Spark entries around 1494 and competitive with Gemini 4 Argon and Claude Opus 5.5 variants. This rapid debut at high rankings, alongside strong results on coding, legal, and memory benchmarks, has anchored trader consensus around the model's initial positioning while leaving room for later refinements or variant releases ahead of the June 2027 resolution window.