MiniCPM5-2B-Abliterated-Uncensored
2BMiniCPMuncensored
from openbmb/MiniCPM5-2B
↓ 11.4K♥ 217 repos
MiniCPM5-2B-heretic-abliterated
MiniCPM5-2B-heretic-abliterated is a Heretic-ablated variant of openbmb/MiniCPM5-2B, published on Hugging Face by insraq. It is 2B parameters, runs on a laptop CPU class and apache-2.0 licence. It is distributed across 8 repositories in GGUF, Transformers, GGUF (imatrix) and MLX formats, totalling 14.1K downloads.
Specification
- Model name
- MiniCPM5-2B-heretic-abliterated
- Source repository
- insraq/MiniCPM5-2B-heretic-abliterated
- Base model
- openbmb/MiniCPM5-2B
- Publisher
- insraq
- Parameters
- 2B
- VRAM required
- ~1.9 GB at Q4_K_M, 4k context (estimated) — runs on 4 GB cards
- Hardware class
- Under 4B — runs on a laptop CPU
- Licence
- apache-2.0
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF, Transformers, GGUF (imatrix), MLX, ONNX
- First indexed
- 7 Sept 2026
- Last updated
- 27 Oct 2026
Ablation technique
How MiniCPM5-2B-heretic-abliterated 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
“We are releasing MiniCPM5-2B, the second model in the MiniCPM5 series, following MiniCPM5-1B. It is a dense 2B Transformer that scales up the same training recipe, built for on-device, local deployment, and resource-constrained scenarios, reaching 2B-class open-source SOTA.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download Abiray/MiniCPM5-2B-heretic-abliterated-GGUF --local-dir ./minicpm5-2b-heretic-abliterated
llama-cli -m ./minicpm5-2b-heretic-abliterated/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "insraq/MiniCPM5-2B-heretic-abliterated"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve insraq/MiniCPM5-2B-heretic-abliterated --trust-remote-codeDownloads and variants (7)
Every published repository of MiniCPM5-2B-heretic-abliterated, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| Abiray/MiniCPM5-2B-heretic-abliterated-GGUF | GGUF | 10.6K |
| mradermacher/MiniCPM5-2B-heretic-abliterated-GGUF | GGUF | 1.2K |
| mradermacher/MiniCPM5-2B-heretic-abliterated-i1-GGUF | GGUF (imatrix) | 858 |
| insraq/MiniCPM5-2B-heretic-abliteratedsource | Transformers | 709 |
| j4ys0n/MiniCPM5-2B-heretic-abliterated-ONNX | ONNX | 381 |
| minte1431/MiniCPM5-2B-heretic-abliterated-GGUF | GGUF | 159 |
| bumblebuttpow/MiniCPM5-2B-heretic-abliterated-MLX-8bit | MLX | 152 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- MiniCPM5-2B-heretic-abliterated 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 openbmb/MiniCPM5-2B, not added by ablation — at 2B parameters, expect a small model’s reasoning and a high error rate on exploit detail.
- 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 openbmb/MiniCPM5-2B’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 MiniCPM5-2B-heretic-abliterated is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with insraq, and does not independently verify publisher claims. See responsible use.