Meta released an updated model, Muse Spark 1.3, in early September, one of a cluster of refreshes across major labs within days of each other. The relentless cadence, several flagship-level updates in a single month, has become the normal rhythm of the AI industry rather than an unusual burst.
For customers, the practical consequence of monthly-scale updates is that capability and price both move quickly, and any snapshot of "the best model" is out of date within weeks.
What it means: do not over-invest in choosing the single best model today, because the ranking will change. Build on tools that make switching models easy and set a recurring reminder, perhaps quarterly, to re-check which option gives the best result for your money.