Meta launched Muse Glimmer on Monday, its first open weights model in more than a year, signaling a return to the open weights arena after critics questioned CEO Mark Zuckerberg's commitment to open source AI following the Llama 4 flop and a restructuring of its AI group. Muse Glimmer is a 30 billion-parameter LLM distilled from Meta's larger, proprietary Muse Spark model. The company describes it as suited for local AI inference workloads, including local agents, code assistants, and applications requiring robust multi-modal tool use and function calling. Released under an Apache 2.0 license, enterprises are free to deploy, use, and modify the model. Early support has hit platforms like Llama.cpp, Ollama, and Unsloth, with optimized implementations expected over the next few weeks. At 30 billion parameters, the model is too small to compete with Chinese models like Moonshot AI's Kimi K3, Alibaba's Qwen 3.8-Max, or DeepSeek V4 Flash. Instead, it targets small-to-medium enterprises and enthusiasts, competing with Alibaba's Qwen 3.6-27B and Google Gemma 4 31B. Meta's benchmarks show Glimmer besting Google's Gemma in most scenarios and trading blows with Alibaba's model. At native BF16 precision, the model fits into a single Nvidia RTX Pro 6000 or AMD MI350P.