Meta Launches Muse Spark 1.3 Multimodal Model

Meta has officially launched Muse Spark 1.3, a proprietary multimodal model that expands the company's growing portfolio of AI capabilities. Officially released on September 2, 2026, the model builds on Meta's earlier work in the Muse family, offering multimodal processing under a proprietary licensing framework.

Release Date and Availability

Muse Spark 1.3 was officially launched on September 2, 2026. The model joins Meta's expanding suite of AI tools, following the earlier Muse Spark 1.1 launch and the separate Meta Muse Glimmer launch. Unlike open-source alternatives from competitors, Muse Spark 1.3 operates under a proprietary license, suggesting commercial deployment may be restricted to Meta's ecosystem and select partners.

Multimodal Capabilities

The model is classified as multimodal, enabling it to process and generate outputs across multiple data types. While specific modality combinations have not been detailed in the initial release documentation, multimodal architectures typically handle combinations of text, images, audio, and video. This positions Muse Spark 1.3 alongside other enterprise-focused multimodal systems in the current AI landscape.

Meta's focus on multimodal processing aligns with industry trends where leading labs are moving beyond text-only systems. The proprietary nature of the license may indicate Meta's intention to integrate the model deeply into its consumer and advertising products rather than offering broad API access.

Muse Spark Series Evolution

Muse Spark 1.3 represents the third documented iteration in the Muse Spark line. The progression from version 1.1 to 1.3 suggests incremental improvements in model architecture, training data, or performance benchmarks, though Meta has not published detailed technical comparisons between versions. The Muse branding appears to encompass multiple model families, including the Glimmer variant released earlier in 2026.

Meta's strategy of iterative versioning mirrors approaches used by other major labs, where point releases deliver measurable performance gains without fundamental architectural redesigns. For organizations evaluating multimodal models, tracking version improvements within a single family offers insights into a lab's development velocity and commitment to ongoing enhancement.

What This Means

Muse Spark 1.3's launch underscores Meta's sustained investment in proprietary AI infrastructure. The proprietary licensing model suggests the company is prioritizing internal applications and controlled partnerships over open research distribution. For developers and enterprises seeking multimodal capabilities, this release expands the competitive landscape but may require direct engagement with Meta for access. As the Muse family continues to evolve, version 1.3 establishes a new baseline for Meta's multimodal performance expectations heading into late 2026.