muse-glimmer-30b
Architecture dense52 LayersLicense apache-2.0Maximum context 128kVocabulary 202,048
At Q4_K_M it is a 15.61 GiB file, and it fits on 12 of the 18 machines we measure (the one with the least memory: Mac with M4 Pro and 24 GB). Here the file is what fills the memory: even at its maximum context of 128k tokens, the context adds 1.70 GiB — and the same content in German takes ×1.30 the tokens of English.
Model size at Q4_K_M
Measured by Local AI Scope
Context memory at 8k tokens
Measured by Local AI Scope
What German costs versus English
Measured by Local AI Scope
Videos: muse-glimmer-30b 3
The best walk-throughs we found. Press play and it opens right here.
The videos live on YouTube, and nothing loads from there until you press play (no-cookie mode). The thumbnail is served by us.
Sizes of the real published files
These are the sizes of the files actually published by the author, added up when the model ships split into parts. Not an estimate from the parameter count.
| Compression | Model size |
|---|---|
Q4_K_M |
15.61 GiB |
This model also reads images. For that it needs a second file, the vision projector (mmproj-Muse-Glimmer-30B-Q4_K_M.gguf, 1.30 GiB), published next to the model and not included in the figures on this page: they count text only. If you are going to send it images, add it to the model size.
Context memory
Computed from the attention pattern this model declares layer by layer, not from the generic formula. That is why the figure is often far smaller than other calculators tell you.
- 8k tokens → 0.18 GiB
- 32k tokens → 0.48 GiB
- 128k tokens → 1.70 GiB
What it costs in each language
Measured by running this model’s own tokenizer over the same content in four languages. Tokens are what you pay for in context, in memory and in your cloud bill.
| Kind of text | EN | ES | FR | DE |
|---|---|---|---|---|
| Code | — | ×1.19 | ×1.32 | ×1.40 |
| Contracts | — | ×1.13 | ×1.23 | ×1.26 |
| Business email | — | ×1.13 | ×1.21 | ×1.23 |
| Support tickets | — | ×1.21 | ×1.37 | ×1.26 |
English is the baseline: every figure is how many times more tokens the same content costs in that language. Average across the whole corpus: DE ×1.30.
Where it fits
Fits means the model, its context memory and the working headroom all fit at the stated context, leaving the system its share. 12 of the 18 machine profiles tested (Q4_K_M, 8k tokens).
Before you hit download: as of 7 October 2026, every program on our list supports this model. llama.cpp needs build b10353 or later, and every stable release since includes it (on 16 October 2026 the latest was v0.7.0, from 15 October 2026); Ollama needs 0.32.8 or later; LM Studio offers it in its own catalogue, which asks for at least 26 GB of system memory; mlx-lm runs it from 0.32.0, text only. vLLM supports it from 0.28.0, but needs the original BF16 weights (55.46 GiB) in graphics-card memory: no PC on our list with a graphics card has that much VRAM, and on a Mac vLLM only runs on the CPU, experimentally. The table counts text only. To send it images you also load the vision projector (1.30 GiB), and the optional DFlash draft model for faster replies adds 1.52 GiB; Ollama loads the projector every time, even for text. With one of the two, the Mac with M4 Pro and 24 GB (21.00 GiB of unified memory for the model) reaches 64k tokens of context instead of 128k, and with both, 32k. The RTX 4090 24 GB (23.20 GiB of VRAM) reaches 128k with both by our count, but with only 0.07 GiB to spare, too little to rely on; at 64k it has 2.08 GiB to spare.