Meta has released Muse Glimmer, a 30B dense multimodal model under the Apache 2.0 license, marking its return to open weights with a focus on "personal superintelligence." The release coincides with Mark Zuckerberg's essay outlining Meta's agenda to empower individuals through AI rather than institutions.

  • Muse Glimmer is optimized for always-on local agents and can be quantized to under 20GB for consumer hardware deployment.
  • The model uses logit distillation from Muse Spark and was trained on agentic traces rather than a conventional base then post-train approach.
  • Third-party analysis places Muse Glimmer at rank 35 on the Intelligence Index, trailing Qwen3.6-27B but scoring well for openness.
  • Meta also promised the upcoming release of Muse Spark 1.2 weights to further support its open-weight frontier strategy.