Meta releases Muse Spark 1.3 with stronger agentic performance
Meta has released Muse Spark 1.3 through Muse Code and the Meta Model API, making it the fourth model in the series introduced within five months. The xhigh tier is available now. The more compute-intensive max tier remains a limited preview for selected partners while additional safety testing continues. Meta also says an open-weights version is planned.
Independent evaluations by Artificial Analysis show the largest improvements on agentic tasks. On its Intelligence Index, max scores 62 points and xhigh 61, compared with 57 for version 1.2. Max leads the current field on τ³-Bench Banking with 52%, while xhigh reaches 47%, matching Claude Fable 5.1 max and GLM-5.3-Flash. On Terminal-Bench 2.1, the scores are 86% and 85%, below Claude Fable 5.1 max at 91.4%.
The model costs $1.25 per million input tokens and $4.25 per million output tokens. Artificial Analysis calculates an index task at $0.55, the lowest price among models scoring at least 59 points. The max price has not been announced. Performance on some tests also declined, including AA-LCR and AA-Omniscience.
Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.