Modelpublic

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

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face

50likes23kdownloads

Model Card

[!IMPORTANT] ## Improvement update — August 15, 2026 This release now includes compact abliterated Q4_K_M companions: a 1.63 GB DFlash drafter and a 1.40 GB multimodal projector, matching Meta's consumer-hardware footprint while preserving this model's modified weights. The complete text quant ladder has also been refreshed with Meta's post-release Jinja correction, which normalizes Reasoning effort to Reasoning strength and prevents duplicate reasoning directives. Text generation, image input, and DFlash speculative decoding were validated together on current llama.cpp.

<h1>MUSE-GLIMMER-30B-ABLITERATED-GGUF</h1>

<h3>GGUF quant ladder of the abliterated Muse Glimmer 30B · runs local on one GPU or CPU</h3>

<p><strong>Built by <a href="https://x.com/Blackfrost_AI">Blackfrost</a> · Las Vegas, NV</strong></p>

</div>

## ✅ All quants live The full text quant ladder (Q2_K → Q8_0), compact Q4KM and full-precision vision projectors, and compact Q4KM and full-precision DFlash drafters are uploaded — see the Files tab.

## ⚠️ EXPERIMENTAL Same-day arch, quantized. Expect sharp edges — decode, coherence, tool-parse, serve edge cases under load. Please open a Community discussion with loader/version, quant, prompt, sampling, and failure mode. Real repros get fixed faster.

Refusal benchmark — R1-HARMFUL-BENCH-450

Measured on the abliterated parent (GGUF quants inherit this behavior):

Metric	Result
True refusal (harmful, n=300)	0 / 300 = 0.0%
True refusal (full 450)	0 / 450 = 0.0%
Substring-harmful	0 / 300
Substring-all	2 / 450 (XSTest false positives)
Errors	0

The weight change removes refusals cleanly — no measured true refusals across the full 450-prompt suite.

Why this model exists

Muse Glimmer is Meta Superintelligence Labs' 30B agentic, on-device model. This is the abliterated build — refusal behavior removed via a Blackfrost weight-change process — packaged as GGUF for llama.cpp, so it runs on a single consumer GPU or CPU, fully offline. The local footprint is the product.

Specifications


Architecture	`muse-glimmer` — dense, 52 layers, hidden 6656, GQA (32 q / 2 kv), sliding-window attention, + vision tower
Base	`meta-models/Muse-Glimmer-30B` — Meta, Apache-2.0
Transform	Abliterated — refusal behavior removed via a Blackfrost weight-change process; multimodal capability intact
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`
Context	131,072
Spec-decode	DFlash drafter — `--spec-type draft-dflash --spec-draft-n-max 15`
Default persona	Ships with the "AI assistant" system template baked in

Quant ladder

quant	size	recommended for
Q2_K	10.0 GB	smallest, quality trade-off
Q3KS	11.7 GB	very tight VRAM
Q3KM	12.7 GB	tight VRAM
Q4KS	15.0 GB	16 GB cards
Q4_K_M	15.8 GB	default — balanced, fits 24 GB
Q5KS	18.0 GB	higher quality
Q5KM	18.5 GB	strong quality/size balance
Q6_K	21.3 GB	near-lossless
Q8_0	27.6 GB	max fidelity

Vision & speculative-decode files

Load a text quant plus an mmproj projector for image input:

file	size	purpose
`mmproj-Muse-Glimmer-30B-Abliterated-Q4_K_M.gguf`	1.40 GB	vision projector — compact, recommended
`mmproj-Muse-Glimmer-30B-Abliterated-F16.gguf`	3.6 GB	vision projector — full precision
`mmproj-Muse-Glimmer-30B-Abliterated-Q8_0.gguf`	1.9 GB	vision projector — compact
`dflash-Muse-Glimmer-30B-Abliterated-Q4_K_M.gguf`	1.63 GB	abliterated DFlash drafter — compact, recommended
`dflash-Muse-Glimmer-30B-Abliterated-F16.gguf`	4.8 GB	DFlash drafter — speculative decoding

Serving (llama.cpp) — confirmed settings

Requires llama.cpp b10353 or newer with llama-server. DFlash runs under llama-server only — it shares the target model's context, so it does not work in llama-cli.

Recommended — with DFlash speculative decoding (~1.6× faster, identical output):

bash

llama-server \
  -m  Muse-Glimmer-30B-Abliterated-Q8_0.gguf \
  -md dflash-Muse-Glimmer-30B-Abliterated-Q4_K_M.gguf \
  --spec-type draft-dflash --spec-draft-n-max 15 \
  -ngl 999 -ngld 999 -fa on --jinja \
  --host 0.0.0.0 --port 8080 -c 16384 \
  --temp 1.0 --top-p 0.95 --top-k 64

—Plain (no drafter): drop -md, --spec-type, --spec-draft-n-max, and -ngld.
—Multimodal (image input): add --mmproj mmproj-Muse-Glimmer-30B-Abliterated-Q4_K_M.gguf.
—One-command kit: `deploy/serve.sh` auto-downloads + serves; full guide in `deploy/DEPLOYMENT.md`.

Confirmed settings

—Sampling: temperature 1.0, top_p 0.95, top_k 64 (Meta). Steer depth with a Reasoning strength: low/medium/high/xhigh system line.
—`--jinja` is required. The refreshed template accepts an OpenAI-style Reasoning effort: <level> line, normalizes it, and does not inject a conflicting second directive.
—Do not stop on `<|eom|>`. Use <|end_of_text|> and <|eot|> as stop tokens.
—`max_tokens` ≥ 1024 — heavy thinker; small budgets return empty `content` because the reasoning channel consumes them. Reasoning arrives in reasoning_content, the answer in content.
—`--spec-draft-n-max 15` — DFlash block size (trained 16, clamped).
—Flash attention: -fa on for peak speed; switch to -fa off if the load hangs on a brand-new GPU paired with an older CUDA toolkit.

Measured performance

1× NVIDIA RTX PRO 6000 (Blackwell), Q8_0, -fa off:

config	decode tok/s	speedup
baseline	~46	1.0×
+ DFlash	~73	1.6×

Speedup rises with -fa on and structured/code output (Meta reports up to 3.1× on an RTX 5090).

<div align="center"> <p>Built by <a href="https://x.com/Blackfrost_AI">Blackfrost</a> · Las Vegas, NV. Not affiliated with Meta.</p> </div>