Muse Glimmer from Meta Superintelligence Labs is now available

Meta's Muse Glimmer, described as the first open model released by Meta Superintelligence Labs, is now available on Ollama, according to an announcement dated August 10, 2026. The model is a 30B multimodal system released under the Apache 2.0 license, purpose-built for agent workloads that run locally, with a context length of 128K or more.

Ollama states that Muse Glimmer can power coding agent applications including Claude Code, Codex, and Pi, as well as long-running personal assistants such as OpenClaw and Hermes. The model can be run with the command ollama run muse-glimmer; on Apple Silicon, Ollama's MLX engine is available through the muse-glimmer:30b-mlx model.

Ollama's launch commands include options for Claude Code, Pi, OpenClaw, and Hermes, and the company says ollama launch also works with Codex, OpenCode, GitHub Copilot and other integrations.

Muse Glimmer supports controllable reasoning strength at low, medium, high, and xhigh levels. Ollama recommends high or xhigh for complex coding and agentic tasks, and lower strengths when speed is a priority.

Ollama's MLX engine now supports DFlash, following earlier multi-token prediction support. According to Ollama, DFlash makes Muse Glimmer run 1.5x to 1.8x faster on Apple Silicon.

Muse Glimmer includes a dedicated 1.8B-parameter perception encoder for native image understanding. With new image input support in the MLX engine, Ollama says coding agents can use it for low-latency, back-to-back tool calling on tasks that require images, including building websites or applications from a drawing or mockup, computer use applications powered by screenshots, and reading documents, receipts, and charts.