Muse Glimmer 30B

Language modelOpen weightsCard summarized

An open model built for always-on local agents.

● Re-checked 12 Oct 2026 · 81 facts · 1 correction · 1 unconfirmedSource: Muse Glimmer Model Card

Use it via API

$255 /month

$5.10 per 1K requests · via Together AI

RAG over docs: 12K in · 600 out · 50K requests/mo · 30% cached · change

Code example Compare

Where to call it

Self-host (weights)meta-models/Muse-Glimmer-30B
Together AImeta-models/Muse-Glimmer-30B

Supports

Image input
Tool calling
Fine-tuning

Full feature support

Good for

Local AI agents: multi-step planning, sequential tool calls and failure recovery on consumer devices
Coding agents for SWE-Bench-style tasks
Schema-based tool and function calling across long multi-turn workflows

Overview

Meta · 02Glimmer

Muse Glimmer

An open model built for always-on local agents.

Cost: $0.35 host
Context: 128K

Cheapest host: $0.35 / $1.50
Max output: —
Knows up to: January 2026
Released: 10 Aug 2026

Muse Glimmer is a 30B dense causal language model with a separate perception encoder, distilled from Muse Spark and built for agentic tasks on consumer hardware. It takes text and images and returns text. Quantized to about 4 bits, the language model is under 20 GB and runs on a 24 GB or 32 GB GPU. Model Card intro; Optimized for Local Deployments

meta-models/Muse-Glimmer-30B

Input / output per 1M: $0.35 / $1.50 via Together AI
Context window: 128K
Max output: —
Knows up to: January 2026
Reasoning: Reasoning strength low / medium / high / xhigh
Weights: Open · Apache 2.0

About this card. A model card is the maker's own account of a model: what it's for, how it was trained and tested, and where it falls short. This page condenses Meta Superintelligence Lab's Muse Glimmer Model Card. Every claim carries its source page.