Zuck rekindles open weights Llama drama with Muse Glimmer

Meta has rekindled its open weights AI ambitions with the launch of Muse Glimmer, a 30 billion-parameter large language model (LLM) unveiled on Monday. Muse Glimmer is a distilled version of Meta's larger proprietary Muse Spark model. This move comes as American tech companies confront the growing influence of Chinese open weights models and call for domestically-created alternatives.

Meta, known for its early investment in open weights model development with the Llama series, has faced criticism regarding its lack of commitment to open source AI after the underwhelming performance of Llama 4 and restructuring of its AI group. With Muse Glimmer, Meta has returned to the open weights arena, positioning the model for small-to-medium sized enterprises and enthusiasts.

Available under a highly permissive Apache 2.0 license, enterprises can freely deploy, use, and modify Muse Glimmer according to their needs. The model is ideally suited for local AI inference workloads, including local agents, code assistants, and applications requiring robust multi-modal tool use and function calling. While not as powerful as some other recent models, Muse Glimmer offers a cost-effective solution for those with limited hardware resources.

Despite its smaller parameter count, Muse Glimmer performs comparably to similarly sized models from Alibaba and Google, according to benchmark tests. However, its performance may not hold up against larger models like Qwen 3.8-27B, which is expected to be released soon. Meta has also committed to releasing an open weights version of its Muse Spark 1.2 model, which competes with OpenAI, Anthropic, and Google. However, the performance of this upcoming model remains uncertain.