# Muse Glimmer — dossier

compiled 2026-10-22 · corpus: 4 non-RT matches (`muse glimmer / muse-glimmer / glimmer 30b / glimmer-30b`), 2026-08-10 → 10-20; pull `_dossiers/_pulls/muse-glimmer.md`.

## Official links

- 2026-08-10 · Introducing Muse Glimmer: An Open Agentic Model That Runs on Your Device — https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model
- weights: https://huggingface.co/meta-models/Muse-Glimmer-30B (linked from the post; model card not retrieved)

## Writings

- 2026-10 · DataCamp (in passing) — https://www.datacamp.com/blog/muse-wonder-and-muse-wonder-full
- not retrieved: secondary explainers (dev.to, mer.vin "first open-weight model since Llama 4", codersera, developersdigest) — some cite Bloomberg naming Muse Spark 1.2 as the distillation parent (unverified)

## Tweets (every line is a record source)

### cited (4)
- 2086912699758227682 @jmbollenbacher 2026-08-10 ♥1 — https://x.com/jmbollenbacher/status/2086912699758227682 — "muse glimmer seems weaker than qwen-3.5. not even 3.6 or 3.8. honestly the qwen 9b models might even be stronger."  [weaker than qwen-3.5]
- 2087110487007731795 @witcheer 2026-08-11 ♥36 · media 1 (0 transcribed) — https://x.com/witcheer/status/2087110487007731795 — "Meta is back in open weights: Muse Glimmer, a dense 30B VLM, Apache 2.0, released yesterday. I benchmarked it on my RTX 5090 28B text decoder plus a 2B visio..."  ['Meta is back in open weights'; local benchmark; image]
- 2109873786044690550 @raunakdoesdev 2026-10-13 ♥5 — https://x.com/raunakdoesdev/status/2109873786044690550 — "changed the default model for my main agent thread in raubot (my personal agent) to muse glimmer! very unexpected subagents still do most of the coding and a..."  [main agent thread default]
- 2112529273273938103 @witcheer 2026-10-20 ♥28 · media 1 (0 transcribed) — https://x.com/witcheer/status/2112529273273938103 — "Muse Glimmer 30B, Meta's new open-weight agent model, runs fully on one RTX 5090 with room to spare, and keeps its speed as the context fills up. I ran Meta'..."  [RTX 5090 speed; image]

### further (0)


### excluded (0)


## tk / open

- HF model card; which Muse Spark version was the teacher (Meta says only "Muse Spark"); character reads