Meta Muse Glimmer: Open-Weight AI for Local Agent Tasks

Meta has released Muse Glimmer, a new open-weight AI model designed for agent tasks, coding, and image input. This model was released in August 2026 under the Apache 2.0 license. Glimmer has about 29.6 billion parameters and uses a dense Transformer design. It can read images, screens, and documents thanks to a 1.8 billion parameter vision encoder. The model supports over 100 languages and has a large 131,072-token context window. Its knowledge cutoff is January 4, 2026. What's interesting is Glimmer's focus on local AI. A 4-bit version can fit into less than 20 gigabytes of hardware. A 17-gigabyte version shows only a 1.0% average score loss across 15 common tests, making it suitable for high-end consumer GPUs. Glimmer also pairs with DFlash, a small decoding model that boosts performance. For example, DFlash can increase speed on an RTX 5090 from 74.9 to 233.4 tokens per second. The bottom line is that Glimmer aims to bring advanced AI agent capabilities to consumer-level hardware.