⌦abliteratedmodels.org

MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE

MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE is a abliterated and uncensored variant of openbmb/MiniCPM-V-4.7-35B-A3B, published on Hugging Face by gbuzhf. It is 35B total parameters (mixture-of-experts), 48GB VRAM class class and apache-2.0 licence. It is distributed across 2 repositories in GGUF format, totalling 21.4K downloads.

Abliteration + uncensored fine-tune35B MoEMiniCPM25B – 50B↓ 21,374♥ 7
View on Hugging Face ↗Base model ↗

Specification

Model name
MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE
Source repository
gbuzhf/MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE
Base model
openbmb/MiniCPM-V-4.7-35B-A3B
Publisher
gbuzhf
Parameters
35B (mixture-of-experts)
VRAM required
~22.1 GB at Q4_K_M, 4k context (estimated) — runs on 24 GB cards
Hardware class
25B – 50B — 48GB VRAM class
Licence
apache-2.0
Task
Text generation
Context window
16K tokens (as documented)
Ablation scope
Not stated by the publisher
Formats available
GGUF
First indexed
14 Oct 2026
Last updated
14 Oct 2026

Ablation technique

How MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE was modified, and what that implies.

Abliteration + uncensored fine-tune

Directional ablation combined with additional fine-tuning on unfiltered data, so behaviour diverges from the base model beyond refusal removal alone.

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

“Five ICE 2.0 tiers of openbmb's MiniCPM-V 4.7 35B-A3B (35B total, 3B active, Qwen3.5-MoE text tower + SigLIP2 vision), uncensored trunk (openbmb's BF16 + Atomic Chat's refusal-direction LoRA, merged), quantized against Sangreal — a purpose-built 77-bucket calibration corpus — and shipping Qwen-Sharp...”

Running it

Commands are templates — confirm the exact repository and quant file before use.

llama.cpp / GGUF
hf download gbuzhf/MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE --local-dir ./minicpm-v-4.7-35b-a3b-uncensored-sangreal-ice
llama-cli -m ./minicpm-v-4.7-35b-a3b-uncensored-sangreal-ice/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "gbuzhf/MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve gbuzhf/MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE --trust-remote-code

Downloads and variants (1)

The only published repository of MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE.

RepositoryFormatDownloads
gbuzhf/MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICEsourceGGUF21.4K

Security-team assessment

Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.

  • MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE 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/MiniCPM-V-4.7-35B-A3B, not added by ablation — at 35B 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.
  • This release adds fine-tuning on unfiltered data, so its behaviour diverges from the base model beyond refusal removal — differences you measure cannot be attributed to ablation alone.
  • Declared licence is apache-2.0, but the terms that bind you are openbmb/MiniCPM-V-4.7-35B-A3B’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 MiniCPM-V-4.7-35B-A3B-Uncensored-Sangreal-ICE is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with gbuzhf, and does not independently verify publisher claims. See responsible use.