Meta ships Muse Spark 1.3 to Muse Code and the Meta Model API, holding max reasoning back for safety testing

Sources: primary, discussion, second thread
Summary: Meta states the model is available today in Muse Code and the Meta Model API, and that the max reasoning mode is coming shortly after it finishes additional safety testing. Every coding figure in the post is Meta's own internal comparison against Muse Spark 1.2, roughly 20 percent fewer tool calls and roughly 25 percent fewer tokens, with no external evaluation cited. Meta states an open weights release for Muse Spark is on the roadmap and gives no date.
Why it matters: The generally available model is not the full model, because the mode Meta rates as needing more safety testing is the one held back, and the efficiency figures a team would plan against are unaudited vendor numbers.
Follow-up: Track when the max reasoning mode is released, whether any external evaluation reproduces the tool call and token reductions, and whether the open weights release gets a date.