Benchmark materials published through Meta channels

The release is accompanied by a benchmark scorecard image posted at Meta's research domain, alongside a supporting image hosted on Meta's media platform. These materials present performance comparisons intended to situate the model — referenced as "Muse Spark Open" — against competing systems.

The claim that the model matches or rivals Chinese counterparts stems from Meta's own framing of the release. The extent to which independent evaluations support that comparison has not been established in the available materials.

Scope of available evidence

Direct access to the primary materials succeeded, and the referenced images were retrieved with their local file hashes recorded, with no access-restricted sources encountered. A total of 8 primary sources were reviewed in connection with the claim. However, an earlier submission associated with the verification effort did not succeed, meaning the documentary record is incomplete.

No independent benchmark results or third-party evaluations of the model's performance relative to Chinese models were available in the reviewed materials. The performance comparison therefore rests on Meta's own published scorecard rather than external testing.

Open questions

The naming and positioning of the model as an open release is consistent with Meta's stated distribution approach, but details such as model size, licensing terms, and availability dates are not specified in the reviewed sources. Readers should treat the "rivals Chinese models" characterization as the company's claim, pending independent corroboration.

The release is real and the materials are accessible, but the competitive performance comparison remains based on Meta's own figures — a reading that is, on balance, Mostly True.