Muse Glimmer 30B pruned V2.5 en pl needs about 17.3 GB of memory at Q4_K_M (a 16.3 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 21 quants, from 7.5 GB (IQ2_XXS) to 54.0 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.7 GB.
Muse Glimmer 30B pruned V2.5 – a 27‑billion‑parameter multimodal LLM for local autonomous agents
Muse Glimmer 30B pruned V2.5 is a 27‑billion‑parameter causal language model with a dedicated perception encoder. It supports reasoning, tool use, and vision on consumer hardware. The model is released under Apache‑2.0 and is available in multiple GGUF quantizations.
Good at
- Multimodal reasoning with text and images
- Reliable tool invocation in long workflows
- End‑to‑end agentic task completion
Limits
- Vision requires a separate mmproj file
- Model is pruned to English + Polish only
- May hang for up to 5 minutes on first start due to tokenizer changes
Best for
- Building local autonomous agents
- Debugging code and writing scripts
- Processing screenshots and documents
Who it’s for. Developers and researchers who need a local, multimodal LLM for agentic tasks without cloud dependency. The model is suitable for systems with limited VRAM and for applications requiring reliable tool use and failure recovery.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XXS | 7.5 GB | 8.5 GB | Too big | Get file → | |
| IQ2_XS | 8.3 GB | 9.2 GB | Too big | Get file → | |
| IQ2_S | 8.7 GB | 9.6 GB | Too big | Get file → | |
| IQ2_M | 9.4 GB | 10.3 GB | Too big | Get file → | |
| Q2_K_S | 9.5 GB | 10.5 GB | Too big | Get file → | |
| Q2_K | 10.2 GB | 11.1 GB | Too big | Get file → | |
| IQ3_XXS | 10.7 GB | 11.6 GB | Too big | Get file → | |
| IQ3_XS | 11.4 GB | 12.4 GB | Too big | Get file → | |
| Q3_K_S | 12.0 GB | 12.9 GB | Too big | Get file → | |
| IQ3_S | 12.0 GB | 12.9 GB | Too big | Get file → | |
| IQ3_M | 12.3 GB | 13.2 GB | Too big | Get file → | |
| Q3_K_M | 13.1 GB | 14.1 GB | Too big | Get file → | |
| Q3_K_L | 14.1 GB | 15.1 GB | Too big | Get file → | |
| IQ4_XS | 14.6 GB | 15.6 GB | Too big | Get file → | |
| IQ4_NL | 15.5 GB | 16.4 GB | Too big | Get file → | |
| Q4_0 | 15.5 GB | 16.4 GB | Too big | Get file → | |
| Q4_K_S | 15.5 GB | 16.5 GB | Too big | Get file → | |
| Q4_K_M | 16.3 GB | 17.3 GB | Too big | Get file → | |
| Q4_1 | 17.1 GB | 18.0 GB | Too big | Get file → | |
| Q8_0 | 28.7 GB | 29.6 GB | Too big | Get file → | |
| BF16 | 54.0 GB | 55.0 GB | Too big | Get file → |
For 8 GB VRAM use Q4_K_M (file size 16.34 GB, estimated 17.3 GB memory). For 12 GB VRAM use Q3_K_M (13.14 GB, 14.1 GB memory). For higher fidelity use Q8_0 (28.71 GB, 29.6 GB memory).
What GPU do you need for Muse Glimmer 30B pruned V2.5 en pl?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: the best quant that fits is IQ3_XXS (10.7 GB file, about 11.6 GB in use).
- 16 GB VRAM: the best quant that fits is IQ4_XS (14.6 GB file, about 15.6 GB in use).
- 24 GB VRAM: the best quant that fits is Q4_1 (17.1 GB file, about 18.0 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (28.7 GB file, about 29.6 GB in use).
- 48 GB VRAM: the best quant that fits is Q8_0 (28.7 GB file, about 29.6 GB in use).
Which Muse Glimmer 30B pruned V2.5 en pl quant should you choose?
Q4_K_M is the usual starting point for Muse Glimmer 30B pruned V2.5 en pl: it balances file size and quality well. Here it is a 16.3 GB file.
The Q8_0 file is 28.7 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is IQ2_XXS at 7.5 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Muse Glimmer 30B pruned V2.5 en pl
How much VRAM does Muse Glimmer 30B pruned V2.5 en pl need?
About 17.3 GB at Q4_K_M with an 8K context window. That is the 16.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Muse Glimmer 30B pruned V2.5 en pl on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 8.5 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Muse Glimmer 30B pruned V2.5 en pl GGUF?
It depends on the quant: IQ2_XXS is 7.5 GB, IQ2_XS is 8.3 GB, IQ2_S is 8.7 GB, IQ2_M is 9.4 GB, Q2_K_S is 9.5 GB, Q2_K is 10.2 GB and 15 more.
What context length does Muse Glimmer 30B pruned V2.5 en pl support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.7 GB at 32K.
Where can I download Muse Glimmer 30B pruned V2.5 en pl GGUF?
The files are on Hugging Face in the bluevoid-pl/Muse-Glimmer-30B-pruned-V2.5-en-pl-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Muse Glimmer 30B pruned V2.5 en pl use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Muse Glimmer 30B pruned V2.5 on 8 GB VRAM?
Yes, using the Q4_K_M quantization requires approximately 17.3 GB of VRAM at 8 k context, which is within 8 GB if using 8 k context and efficient batching.
Do I need a separate vision model?
Yes, vision requires downloading the mmproj file. Use the Q8_0 mmproj for best compatibility.
Will the model work with Polish prompts?
The model is pruned to English + Polish characters, but it performs best with English prompts; Polish may reduce performance.