---
title: Muse Glimmer 30B
emoji: ✨
colorFrom: purple
colorTo: indigo
sdk: static
app_file: index.html
short_description: Chat with Meta's Muse Glimmer-30B multimodal model
python_version: "3.12"
startup_duration_timeout: 1h
models:
  - meta-models/Muse-Glimmer-30B
license: apache-2.0
---

# Muse Glimmer-30B Demo

> **Status:** awaiting a ZeroGPU community grant. The Gradio app (`app.py`) is complete and
> ZeroGPU-ready. Once the grant lands, flip the frontmatter to
> `sdk: gradio`, `sdk_version: 6.22.0`, `app_file: app.py`.

Multimodal chat demo for [meta-models/Muse-Glimmer-30B](https://huggingface.co/meta-models/Muse-Glimmer-30B),
Meta Superintelligence Lab's 30B agentic model (Apache 2.0).

- **Image + text input** — drop in screenshots, charts, or documents alongside your prompt.
- **Visible reasoning** — the model's `to=self` reasoning channel streams into a collapsible block before the final answer.
- **Reasoning strength** — pick low / medium / high / xhigh via the chat template's native control.
- **Quantized to NF4** (~17 GB) with bitsandbytes so it fits ZeroGPU's 48 GB slice; the model card reports ≤1% degradation at 4-bit.
- Sampling defaults follow the model card: temperature 1.0, top-p 0.95, top-k 64.

Also exposed as an **MCP server** — point your agent at this Space's `/gradio_api/mcp/` endpoint.