Meta open-sources 816B-parameter Muse Spark for coding, agents
Meta Platforms has released open weights for its 816-billion-parameter Muse Spark model, previously proprietary since its April debut. The model, called Muse Spark Open, is designed for coding and long-running agent tasks.
The release follows Meta’s October launch of Muse Wonder models, which also focused on agent workflows. It arrives as open-source AI tools see growing interest among developers and enterprises. Some industry observers suggest models like Muse Spark Open could provide options for teams working outside closed-agent platforms, though adoption will depend on factors like cloud provider support and integration ease.
The 816-billion-parameter scale of Muse Spark Open is larger than some recent open releases, such as AWS’s 2-billion-parameter Strands Decider 2B and Nvidia’s Open Agent Safety Platform. However, the practical value of these models varies by use case.
Meta’s decision to open Muse Spark’s weights comes amid ongoing discussions about AI accessibility. Some startups advocate for open coding tools as a way to avoid vendor lock-in, while others highlight risks associated with agent autonomy. Whether Muse Spark Open gains traction will likely hinge on how developers and enterprises assess its capabilities against existing options.
For now, the release highlights Meta’s role in the evolving open AI space, alongside efforts from other companies like AWS and Nvidia.
Sources: siliconangle.com
“Meta’s release of Muse Spark Open adds another high-capacity model to the open-weight AI landscape, potentially offering developers an alternative as debates over proprietary versus open systems continue.”
Read the original reporting
The outlets below did the original reporting.
- Meta releases 816B-parameter Muse Spark Open for coding and agent work — siliconangle.com
Related briefs
This brief was drafted automatically from the sources above and published under our editorial policy. Spotted an error? Tell us.