Overview
- Meta began rolling out Muse Spark 1.3 through Muse Code and the Meta Model API on Wednesday, with the highest‑reasoning “max” mode kept in limited preview while the company finishes additional safety checks.
- Meta reports Muse Spark 1.3 uses about 20% fewer external tool calls and about 25% fewer tokens than version 1.2, and the company says the model better sustains long, multi‑step agentic and coding workflows.
- Independent benchmark firm Artificial Analysis placed Muse Spark 1.3 near the top of evaluated models, and Meta’s AI leadership publicly highlighted those results in comparisons with rival models.
- The release underscores Meta’s strategic shift from freely distributed Llama weights to a two‑track approach that pairs closed, paid frontier models with selective smaller open‑weight releases such as Muse Glimmer.
- For developers and enterprises, the update promises lower runtime costs and tighter agent capabilities through Muse Code, but safety gating of the highest reasoning levels will limit full access while Meta completes further testing.