Meta has released Muse Glimmer, a 30-billion-parameter open-weight model built for agentic tasks that can run locally on a consumer device with a single GPU, and will soon release the weights for Muse Spark 1.2, its most capable model yet. Muse Glimmer is released under the Apache 2.0 licence, with weights available through Hugging Face, and is intended for local agents, function calling, coding, and LLM-as-a-judge workloads without requiring a cloud connection. The company hopes to replicate the success of its Llama series, which allowed developers to fine-tune, quantise, and deploy models on their own infrastructure, but Llama 4, launched a year ago, failed to generate the same momentum. Meta subsequently overhauled its AI efforts, bringing Alexandr Wang to lead Meta Superintelligence Labs, which launched Muse Spark in April this year. The release comes as the developer community seeks frontier open-weights to bring down token costs. Adithya S Kolavi, an AI researcher at Hugging Face, called it a "Huge win for the open ecosystem." Analyst Kyle Chan believes an open-weight Muse Spark 1.2 would offer similar capabilities to top Chinese models like Kimi K3 and GLM-5.2, and that Meta’s US origins could give it an advantage among Western users. Amanda Brock, CEO of OpenUK, noted the models are shared on a true open source licence without commercial restrictions,