Model · Meta · Text
Muse Glimmer 30B
Meta's current open-weight model and the practical successor to the Llama line for local deployment. An agentic model built for tool use and multi-step task completion, and Meta's first language model under a standard OSI-approved licence.
- Modality
- Text
- License
- Apache 2.0 (Open)
- Parameter size
- ~30B dense (including a ~1.8B vision encoder)
- Context window
- 131,072 tokens 131,072+ tokens. Published under the meta-models Hugging Face org, not meta-llama.
- Released
- August 10, 2026
- Last verified
- October 28, 2026
- Runs locally
- Yes
- Also handles
- vision, reasoning
What you need to run it
- Mac
- Yes. Any Apple Silicon Mac (M1 or newer) with at least 24 GB (32 GB recommended) of memory. Newer chips are faster, but memory is what decides whether it runs.
- Windows / Linux PC
- NVIDIA works best. A graphics card with at least 24 GB (32 GB recommended) of video memory.
- No graphics card
- Not practical. Too slow without a graphics card or Apple Silicon.
Meta ships ready-made 4-bit builds for 24 GB and 32 GB machines and measured it on Apple M4 and M5 Max MacBooks.
Numbers are for: 4-bit build of the 30B model (about 17-20 GB). Based on the maker’s own figures; where the maker gives only a graphics-card figure, the Mac memory is our estimate.
Strengths
- Apache 2.0 — Meta's first language model under a standard OSI-approved licence rather than the bespoke Llama Community License
- Runs on a single 24GB+ consumer GPU, filling the local-deployment slot Llama 4 left empty
- Built end-to-end for agentic work: reliable tool use and function calling across extended workflows
- Multimodal text and image, with support for 100+ languages
Weaknesses
- Not optimised for video — video input is processed as individual frames
- Can still make errors in multi-step reasoning, and may produce inaccurate or biased responses
- Performance degrades on languages outside the strongly supported set
- No audio input or output
Try it
| Where | Type | Notes |
|---|---|---|
| Hugging Face | weights | Apache 2.0; GGUF and ExecuTorch builds also published |
Used in solutions
Official sources
- Model card model
- Announcement announcement
Change log
- — Added hardware requirements (Mac support, memory, NVIDIA). Source: Meta model card meta-models/Muse-Glimmer-30B (24/32 GB targets).
- — Verified current — Meta's launch post still lists Muse Glimmer as its open-weight model under Apache 2.0. No newer Glimmer version; the other Meta models the post lists (Muse Spark 1.1 and 1.2, Muse Image) are separate lines.
- — Initial entry. Released 2026-08-10 under the new meta-models Hugging Face org. Meta's open-weight work has moved off the Llama name — there is no Llama 5, and the closed-weight Muse Spark (2026-04-08) replaced Llama as Meta's flagship. Llama 3.1 8B is kept in the catalog because Llama 4's smallest model is 109B, leaving Muse Glimmer as the only current Meta option for a single consumer GPU.
Esc