Meta’s rapid iteration on the Muse Spark family drives current trader sentiment, with version 1.3 shipping September 2 after roughly four-week gaps from 1.2 and 1.1. The multimodal large language model emphasizes agentic workflows, long-horizon coding, tool use, and efficiency gains—such as 20-25% fewer tokens and tool calls—while supporting a 1M-token context window through the Meta Model API and Muse Code. Recent benchmarks show strong results on ProgramBench and similar evaluations, positioning it competitively against frontier systems despite closed weights. Prediction markets currently imply a late-October median release for 1.4+, with roughly 70% probability by month-end, reflecting Meta Superintelligence Labs’ consistent cadence and focus on real-world usability ahead of potential developer events or API updates.