Meta 發布適用於本地硬體的 Muse AI 模型
Meta Releases Muse AI Models for Local Hardware
Updated at: August 11, 2026 at 04:15 AM
在2026年8月10日，Meta在人工智慧領域邁出重要一步，發布了Muse Glimmer。
On August 10, 2026, Meta made a significant move in the AI landscape by releasing Muse Glimmer, a 30-billion-parameter open-weight model designed for local operation.
這是一款擁有300億參數、專為在地端運行而設計的開放權重模型。
Unlike standard chatbots, this model is built for autonomous, long-running agentic tasks such as software automation and complex file management.
與標準聊天機器人不同，該模型專為自動化、長時間運行的代理任務所打造。
By using a dense architecture, Muse Glimmer ensures reliability and consistent instruction following.
透過使用稠密架構，Muse Glimmer確保了可靠性與一致的指令遵循能力。
One of its standout features is the ability to process data directly on personal hardware, which enhances privacy by ensuring sensitive information never leaves the user’s device.
其最突出的特點之一是能夠直接在個人硬體上處理數據，藉此確保敏感資訊不會離開使用者的裝置，進而增強隱私性。
It is optimized for high-end consumer GPUs and supports a massive 120K+ token context window, alongside multimodal capabilities like interpreting charts and screenshots.
它針對高階消費級顯示卡(GPU)進行了優化，支援超過12萬個標記(token)的超大上下文視窗，並具備解讀圖表和螢幕截圖等的多模態能力。
Released under the permissive Apache 2.0 license, Muse Glimmer is free for commercial and personal use, marking a return to Meta's open-source roots.
Muse Glimmer採用寬鬆的Apache 2.0授權釋出，可免費供商業與個人使用，這標誌著Meta回歸其開源根基。
This model, distilled from the larger Muse Spark system, offers high performance for tasks requiring deep reasoning.
該模型源自更龐大的Muse Spark系統，對於需要深度推理的任務提供高效能表現。
With broad compatibility across tools like Ollama and LM Studio, Meta is positioning Muse Glimmer to challenge both cloud-based proprietary models and other open-weight competitors, signaling an exciting shift toward accessible, high-powered local AI.
透過對Ollama和LM Studio等工具的廣泛相容性，Meta將Muse Glimmer定位為挑戰雲端專有模型及其他開放權重競爭對手的利器，象徵著朝向易於取得、高功率在地人工智慧邁進的令人振奮的轉變。
