Hugging Face's logo

Blackfrost-AI
/
Muse-Glimmer-30B-Abliterated-GGUF

Image-Text-to-Text
GGUF
muse-glimmer
abliterated
quantized
agentic
multimodal
llama.cpp
dflash
experimental
conversational
Model card Files
xet
 Community
1
Blackfrost-AI commited on
Commit
64052f4
·
0 Parent(s):

Collapse history — purge BF16 master reference from all commits

Browse files
Files changed (14) hide show
  1. .gitattributes +48 -0
  2. Muse-Glimmer-30B-Abliterated-Q2_K.gguf +3 -0
  3. Muse-Glimmer-30B-Abliterated-Q3_K_M.gguf +3 -0
  4. Muse-Glimmer-30B-Abliterated-Q3_K_S.gguf +3 -0
  5. Muse-Glimmer-30B-Abliterated-Q4_K_M.gguf +3 -0
  6. Muse-Glimmer-30B-Abliterated-Q4_K_S.gguf +3 -0
  7. Muse-Glimmer-30B-Abliterated-Q5_K_M.gguf +3 -0
  8. Muse-Glimmer-30B-Abliterated-Q5_K_S.gguf +3 -0
  9. Muse-Glimmer-30B-Abliterated-Q6_K.gguf +3 -0
  10. Muse-Glimmer-30B-Abliterated-Q8_0.gguf +3 -0
  11. README.md +131 -0
  12. dflash-Muse-Glimmer-30B-Abliterated-F16.gguf +3 -0
  13. mmproj-Muse-Glimmer-30B-Abliterated-F16.gguf +3 -0
  14. mmproj-Muse-Glimmer-30B-Abliterated-Q8_0.gguf +3 -0
.gitattributes ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ mmproj-Muse-Glimmer-30B-Abliterated-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
37
+ mmproj-Muse-Glimmer-30B-Abliterated-F16.gguf filter=lfs diff=lfs merge=lfs -text
38
+ dflash-Muse-Glimmer-30B-Abliterated-F16.gguf filter=lfs diff=lfs merge=lfs -text
39
+ Muse-Glimmer-30B-Abliterated-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
40
+ Muse-Glimmer-30B-Abliterated-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
41
+ Muse-Glimmer-30B-Abliterated-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
42
+ Muse-Glimmer-30B-Abliterated-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
43
+ Muse-Glimmer-30B-Abliterated-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
44
+ Muse-Glimmer-30B-Abliterated-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
45
+ Muse-Glimmer-30B-Abliterated-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
46
+ Muse-Glimmer-30B-Abliterated-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
47
+ Muse-Glimmer-30B-Abliterated-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
48
+ Muse-Glimmer-30B-Abliterated-BF16.gguf filter=lfs diff=lfs merge=lfs -text
Muse-Glimmer-30B-Abliterated-Q2_K.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb0815ebc60fe168d6d397b6b8c0d37dadfe556519ee8471a224fe1e0c9082fc
3
+ size 10686701184
Muse-Glimmer-30B-Abliterated-Q3_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0368edbe307d27fa8b63697fac9ef23ae9008451c6c2a201695b387c82f3222d
3
+ size 13680520064
Muse-Glimmer-30B-Abliterated-Q3_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36e80efb37948879736240b33f578dab760b9e3d8e9d00092ad2d84a1d4c548c
3
+ size 12512312192
Muse-Glimmer-30B-Abliterated-Q4_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b2e7e4ee0956f0c1db809a74f688775a0e9b1a050d6b90b38c08571ea57fc0a
3
+ size 16935294080
Muse-Glimmer-30B-Abliterated-Q4_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbd5057a21abb792fc92ca9daf3d1677b4bac5de7d11ab4d514d5a0681eed63e
3
+ size 16133512320
Muse-Glimmer-30B-Abliterated-Q5_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:551a71ca3a0745ab65140ee9b44b9e0b2ef582446b242f0515e50203e66e4e8f
3
+ size 19811378304
Muse-Glimmer-30B-Abliterated-Q5_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89406eefdd40477b7c5ae7576bb0a9999a334fcb457605b749827536f0b0a26f
3
+ size 19346549888
Muse-Glimmer-30B-Abliterated-Q6_K.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9b59505861f7521355028bd56a2c9b8ae739e0b74bf338f5c8a4b7c4b53bbd4
3
+ size 22867217792
Muse-Glimmer-30B-Abliterated-Q8_0.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:313e334f4a57fbd74a1e949e70cea8824d0a3d3091c1de485d4ba2903f762a13
3
+ size 29612960640
README.md ADDED
@@ -0,0 +1,131 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: meta-models/Muse-Glimmer-30B
4
+ tags:
5
+ - muse-glimmer
6
+ - gguf
7
+ - abliterated
8
+ - quantized
9
+ - agentic
10
+ - multimodal
11
+ - llama.cpp
12
+ - dflash
13
+ - experimental
14
+ pipeline_tag: image-text-to-text
15
+ library_name: gguf
16
+ ---
17
+
18
+ <div align="center">
19
+
20
+ <h1>MUSE-GLIMMER-30B-ABLITERATED-GGUF</h1>
21
+
22
+ <h3>GGUF quant ladder of the abliterated Muse Glimmer 30B · runs local on one GPU or CPU</h3>
23
+
24
+ <p><strong>Built by <a href="https://x.com/Blackfrost_AI">Blackfrost</a> · Las Vegas, NV</strong></p>
25
+
26
+ <p>
27
+ <img src="https://img.shields.io/badge/GGUF_full_ladder-047857?style=for-the-badge" />
28
+ <img src="https://img.shields.io/badge/0%2F450_refusals-047857?style=for-the-badge" />
29
+ <img src="https://img.shields.io/badge/Abliterated-1f2937?style=for-the-badge" />
30
+ <img src="https://img.shields.io/badge/EXPERIMENTAL-b45309?style=for-the-badge" />
31
+ <img src="https://img.shields.io/badge/llama.cpp_+_DFlash-1f2937?style=for-the-badge" />
32
+ </p>
33
+
34
+ </div>
35
+
36
+ > ## ✅ All quants live
37
+ >
38
+ > The full text quant ladder (`Q2_K` → `Q8_0`), the vision `mmproj` projector, and the DFlash drafter are all uploaded — see the **Files** tab.
39
+
40
+ > ## ⚠️ EXPERIMENTAL
41
+ >
42
+ > Same-day arch, quantized. Expect sharp edges — decode, coherence, tool-parse, serve edge cases under load. Please open a [Community discussion](https://huggingface.co/Blackfrost-Research/Muse-Glimmer-30B-Abliterated-GGUF/discussions) with loader/version, quant, prompt, sampling, and failure mode. Real repros get fixed faster.
43
+
44
+ ---
45
+
46
+ ## Refusal benchmark — R1-HARMFUL-BENCH-450
47
+
48
+ Measured on the abliterated parent (GGUF quants inherit this behavior):
49
+
50
+ | Metric | Result |
51
+ |---|--:|
52
+ | True refusal (harmful, n=300) | **0 / 300 = 0.0%** |
53
+ | True refusal (full 450) | **0 / 450 = 0.0%** |
54
+ | Substring-harmful | 0 / 300 |
55
+ | Substring-all | 2 / 450 (XSTest false positives) |
56
+ | Errors | 0 |
57
+
58
+ The in-place weight change removes the refusal direction cleanly with no measured true refusals across the full 450-prompt suite.
59
+
60
+ ---
61
+
62
+ ## Why this model exists
63
+
64
+ Muse Glimmer is Meta Superintelligence Labs' 30B **agentic, on-device** model. This is the **abliterated** build — the refusal direction removed via an in-place residual-write weight change — packaged as **GGUF for llama.cpp**, so it runs on a single consumer GPU or CPU, fully offline. The local footprint is the product.
65
+
66
+ ---
67
+
68
+ ## Specifications
69
+
70
+ | | |
71
+ |---|---|
72
+ | **Architecture** | `muse_glimmer` — dense, 52 layers, hidden 6656, GQA (32 q / 2 kv), sliding-window attention, + vision tower |
73
+ | **Base** | [`meta-models/Muse-Glimmer-30B`](https://huggingface.co/meta-models/Muse-Glimmer-30B) — Meta, Apache-2.0 |
74
+ | **Transform** | Abliteration only — in-place residual-write weight change (attn `o_proj` + `mlp.down_proj`), α=1.5 × 3 iterative passes. Vision / gates / norms untouched. |
75
+ | **Formats** | GGUF — `Q2_K`, `Q3_K_S`, `Q3_K_M`, `Q4_K_S`, `Q4_K_M`, `Q5_K_S`, `Q5_K_M`, `Q6_K`, `Q8_0` |
76
+ | **Context** | 131,072 |
77
+ | **Spec-decode** | DFlash drafter — `--spec-type draft-dflash --spec-draft-n-max 15` |
78
+ | **Default persona** | Ships with the "AI assistant" system template baked in |
79
+
80
+ ---
81
+
82
+ ## Quant ladder
83
+
84
+ | quant | size | recommended for |
85
+ |---|--:|---|
86
+ | Q2_K | 10.0 GB | smallest, quality trade-off |
87
+ | Q3_K_S | 11.7 GB | very tight VRAM |
88
+ | Q3_K_M | 12.7 GB | tight VRAM |
89
+ | Q4_K_S | 15.0 GB | 16 GB cards |
90
+ | **Q4_K_M** | **15.8 GB** | **default — balanced, fits 24 GB** |
91
+ | Q5_K_S | 18.0 GB | higher quality |
92
+ | Q5_K_M | 18.5 GB | strong quality/size balance |
93
+ | Q6_K | 21.3 GB | near-lossless |
94
+ | Q8_0 | 27.6 GB | max fidelity |
95
+
96
+ ---
97
+
98
+ ## Vision & speculative-decode files
99
+
100
+ Load a text quant **plus** an `mmproj` projector for image input:
101
+
102
+ | file | size | purpose |
103
+ |---|--:|---|
104
+ | `mmproj-Muse-Glimmer-30B-Abliterated-F16.gguf` | 3.6 GB | vision projector — full precision |
105
+ | `mmproj-Muse-Glimmer-30B-Abliterated-Q8_0.gguf` | 1.9 GB | vision projector — compact |
106
+ | `dflash-Muse-Glimmer-30B-Abliterated-F16.gguf` | 4.8 GB | DFlash drafter — speculative decoding |
107
+
108
+ ---
109
+
110
+ ## Serving (llama.cpp / LM Studio)
111
+
112
+ ```bash
113
+ # text:
114
+ llama-server -hf Blackfrost-Research/Muse-Glimmer-30B-Abliterated-GGUF:Q4_K_M --ctx-size 32768
115
+
116
+ # multimodal (text quant + vision projector):
117
+ llama-server -hf Blackfrost-Research/Muse-Glimmer-30B-Abliterated-GGUF:Q4_K_M \
118
+ --mmproj mmproj-Muse-Glimmer-30B-Abliterated-F16.gguf
119
+
120
+ # with DFlash speculative decoding:
121
+ llama-server -hf Blackfrost-Research/Muse-Glimmer-30B-Abliterated-GGUF:Q4_K_M \
122
+ --spec-type draft-dflash --spec-draft-n-max 15
123
+ ```
124
+
125
+ **Sampling (Meta guidance):** `temperature=1.0, top_p=0.95, top_k=64`. Steer reasoning depth with a `Reasoning strength: low/medium/high/xhigh` system line. Budget a generous `max_tokens` — it's a heavy thinker.
126
+
127
+ ---
128
+
129
+ <div align="center">
130
+ <p>Built by <a href="https://x.com/Blackfrost_AI">Blackfrost</a> · Las Vegas, NV. Not affiliated with Meta.</p>
131
+ </div>
dflash-Muse-Glimmer-30B-Abliterated-F16.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8356491407f97cddcef17e66f4f05ed3be4e8b7cc2777eb32f0ea16296d031c2
3
+ size 5125208544
mmproj-Muse-Glimmer-30B-Abliterated-F16.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2e08b8bd13ca60b59ffdcb79b969b354b1edda9b9198047d10621adc11591bb
3
+ size 3847367616
mmproj-Muse-Glimmer-30B-Abliterated-Q8_0.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca930c0d8b0ee0fa1a569c8ac05d090ff90440aaad6f2ebcc9bdec1433b32d66
3
+ size 2049878976