Meta’s Muse Spark series from its Superintelligence Labs has followed a rapid release cadence, with version 1.3 launching on September 2, 2026, via the Meta Model API and Muse Code agent. That update emphasized longer-horizon agentic workflows, coding efficiency (roughly 20% fewer tool calls and 25% fewer tokens than 1.2), and a 1-million-token context window while maintaining multimodal capabilities. Earlier iterations like the original Muse Spark and 1.1 entered Arena leaderboards with competitive but not always frontier-leading text scores (e.g., around 1472 on Pareto metrics for 1.1), reflecting Meta’s balance between product usability and benchmark optimization. The next 1.4+ release will likely target further gains in agentic performance and coding to close gaps with models from OpenAI, Anthropic, and Google, though the exact timing, feature priorities, and whether the initial public leaderboard appearance clears high thresholds like 1480 remain uncertain. Traders monitor Meta’s API expansions, any open-weights signals, and internal benchmark leaks as key signals ahead of the June 2027 resolution.