darkc0de_Muse-Glimmer-30B-heretic
30BOtherheretic
from darkc0de/Muse-Glimmer-30B-heretic
↓ 25.4K♥ 15
Homura-30B
Homura-30B is a Heretic-ablated variant of darkc0de/Muse-Glimmer-30B-heretic, published on Hugging Face by hyrelabs. It is 30B parameters, 48GB VRAM class class and apache-2.0 licence. It is distributed across 4 repositories in GGUF and MLX formats, totalling 68.8K downloads.
Specification
- Model name
- Homura-30B
- Source repository
- hyrelabs/Homura-30B-GGUF
- Base model
- darkc0de/Muse-Glimmer-30B-heretic
- Publisher
- hyrelabs
- Parameters
- 30B
- Hardware class
- 25B – 50B — 48GB VRAM class
- Licence
- apache-2.0
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF, MLX
- First indexed
- 13 Aug 2026
- Last updated
- 23 Aug 2026
Ablation technique
How Homura-30B was modified, and what that implies.
Heretic
Automated directional ablation via the Heretic toolchain, which searches for the refusal direction and applies it with a KL-divergence budget so general capability is preserved.
Reported scope: Not stated by the publisher. Publishers frequently omit this, so verify behaviour empirically rather than assuming full coverage.
From the publisher’s model card
“HYRE's first in-house model — an agent-tuned, uncensored derivative of Meta's Muse Glimmer 30B, built for autonomous agents that need tool-calling and a straight-talking voice with no refusal walls.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download hyrelabs/Homura-30B-GGUF --local-dir ./homura-30b
llama-cli -m ./homura-30b/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "hyrelabs/Homura-30B-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve hyrelabs/Homura-30B-GGUF --trust-remote-codeDownloads and variants (3)
Every published repository of Homura-30B, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| hyrelabs/Homura-30B-GGUFsource | GGUF | 68.7K |
| 3zhil/Homura-30B-GGUF | GGUF | 71 |
| hyrelabs/Homura-30B-MLX-4bit | MLX | 45 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Homura-30B will attempt offensive-security prompts that a hosted commercial model declines, which is what makes it usable for red-team corpus generation and for measuring what an unaligned model of this class produces.
- Capability is inherited from darkc0de/Muse-Glimmer-30B-heretic, not added by ablation — at 30B parameters, expect roughly the base model’s competence, minus some instruction-following fidelity.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- Declared licence is apache-2.0, but the terms that bind you are darkc0de/Muse-Glimmer-30B-heretic’s — a re-release cannot grant more than its parent.
Handling: run it isolated, put moderation in front of it if anyone outside your team can reach it, and verify the weights before loading — these are third-party artifacts. Full handling and licence guidance.
Metadata for Homura-30B is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with hyrelabs, and does not independently verify publisher claims. See responsible use.