Meta's New Open Multimodal Model Muse Glimmer Lands on Ollama

ollama · x · 2026-08-10

Meta Superintelligence Labs has released Muse Glimmer (30B), its newest open model now available on Ollama. Purpose-built for local agent workloads, the multimodal model supports image input and is optimized for Apple Silicon via Ollama's MLX engine.

Leveraging the new DFlash technology alongside multi-token prediction (MTP), the model runs 1.5×–1.8× faster on Mac hardware.

Developers can seamlessly integrate Muse Glimmer into popular coding agents (like Claude Code, Codex, GitHub Copilot) or personal assistant frameworks (like OpenClaw and Hermes) using simple CLI commands. It also features controllable reasoning strengths (low, medium, high, xhigh) to balance speed and complexity for agentic tasks.

Related event: Meta Open-Sources Muse Glimmer: A 30B Agent Model for Consumer GPUs(30 posts)→

Original post →

More from coding & agent

coding & agent channel →