Muse Glimmer 30B, GGUF
Muse Glimmer 30B is a multimodal dense causal language model designed for autonomous agentic tasks, combining multi-step reasoning, reliable tool use, failure recovery, and visual understanding. The language model uses a repeating local-to-global attention pattern and GQA. It supports text and image inputs and is optimized for local deployment and long-horizon agent workflows.
This repository holds Vultr's GGUF quantizations of it. They are hybrid quantizations: rather than applying one precision to the whole network, each layer's precision is chosen from a measured importance score, so the bits go where they change the output most.
Model Overview
| Attribute | Value |
|---|---|
| Base model | meta-models/Muse-Glimmer-30B |
| Architecture | muse-glimmer, dense |
| Parameters | 30B |
| Perception encoder parameters | 1.8B |
| Layers | 52 |
| Hidden size | 6,656 |
| Attention heads | 32 |
| Query heads | 32 |
| Key-value heads | 2 |
| Context length | 131K+ tokens |
| Vocabulary | 202,048 |
| Attention pattern | 3:1 local-to-global, 2,048-token sliding window |
| License | Apache 2.0 |
| Attribute | Value |
|---|---|
| Base model | meta-models/Muse-Glimmer-30B |
| Architecture | muse-glimmer, dense |
| Parameters | 30B |
| Perception encoder parameters | 1.8B |
| Layers | 52 |
| Hidden size | 6,656 |
| Attention heads | 32 |
| Query heads | 32 |
| Key-value heads | 2 |
| Context length | 131K+ tokens |
| Vocabulary | 202,048 |
| Attention pattern | 3:1 local-to-global, 2,048-token sliding window |
| License | Apache 2.0 |
Quantizations
| File | Type | Size | KLD vs bf16 |
|---|---|---|---|
| Muse-Glimmer-30B-Q5_K_M.gguf | Q5_K_M | 21.79 GB | 0.0040 |
| Muse-Glimmer-30B-Q4_K_M.gguf | Q4_K_M | 18.63 GB | 0.0112 |
| Muse-Glimmer-30B-IQ3_M.gguf | IQ3_M | 14.10 GB | 0.0430 |
| Muse-Glimmer-30B-IQ3_S.gguf | IQ3_S | 13.77 GB | 0.0482 |
| File | Type | Size | KLD vs bf16 |
|---|---|---|---|
| Muse-Glimmer-30B-Q5_K_M.gguf | Q5_K_M | 21.79 GB | 0.0040 |
| Muse-Glimmer-30B-Q4_K_M.gguf | Q4_K_M | 18.63 GB | 0.0112 |
| Muse-Glimmer-30B-IQ3_M.gguf | IQ3_M | 14.10 GB | 0.0430 |
| Muse-Glimmer-30B-IQ3_S.gguf | IQ3_S | 13.77 GB | 0.0482 |
The vision encoder is in Muse-Glimmer-30B-mmproj-f16.gguf (3.85 GB). llama.cpp downloads it automatically with the command under Run, so image input works without extra steps.
KL divergence measures how far a quantized model's next-token probabilities drift from those of the original bf16 model, averaged over every token of the evaluation text. The lower the value, the closer the quantized model is to the original; zero would mean identical output.
Evaluation
Every variant in this repository was evaluated against the original bf16 model for perplexity, KL divergence, top-token agreement, HellaSwag and WinoGrande, under identical conditions. The full results are maintained on the model's page in the Vultr Model Library:
https://docs.vultr.com/models/meta-muse-glimmer/muse-glimmer-30b?tab=quantized
That page is the reference for these files. The summary above is an excerpt from it.
Run
llama-server -hf vultr/Muse-Glimmer-30B-GGUF:Q4_K_M
The model's chat template is embedded in every file, so no extra prompt formatting is needed.
License
These files are derived from meta-models/Muse-Glimmer-30B and are distributed under the same license, Apache 2.0. The model is the work of Meta Inc.; Vultr produced the quantizations.