⌦abliteratedmodels.org

NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated

NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated is a refusal-ablated variant of nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16, published on Hugging Face by BoldingBuilds. It is 74B total parameters (mixture-of-experts), multi-GPU class and other licence. It is distributed across 2 repositories in GGUF format, totalling 9.5K downloads.

Abliteration (directional ablation)74B MoENemotron50B – 100B↓ 9,539♥ 12
View on Hugging Face ↗Base model ↗

Specification

Model name
NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated
Source repository
BoldingBuilds/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated-24GB-GGUF
Base model
nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16
Publisher
BoldingBuilds
Parameters
74B (mixture-of-experts)
VRAM required
~45.7 GB at Q4_K_M, 4k context (estimated) — runs on 48 GB cards
Hardware class
50B – 100B — multi-GPU
Licence
other
Task
Image text to text
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
GGUF
First indexed
16 Oct 2026
Last updated
16 Oct 2026

Ablation technique

How NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated was modified, and what that implies.

Abliteration (directional ablation)

The single residual-stream direction that mediates refusal is identified from harmful/harmless prompt pairs and projected out of the model weights.

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

“An uncensored edit of NVIDIA Nemotron-3.5-Bolt-74B-A9B, a hybrid Mamba/MoE model, produced by editing NVIDIA's shipped 24GB 2-bit GGUF (nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-GGUF) directly — no BF16 requantize at any point.”

Running it

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

llama.cpp / GGUF
hf download BoldingBuilds/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated-24GB-GGUF --local-dir ./nvidia-nemotron-3.5-bolt-74b-a9b-abliterated
llama-cli -m ./nvidia-nemotron-3.5-bolt-74b-a9b-abliterated/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "BoldingBuilds/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated-24GB-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve BoldingBuilds/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated-24GB-GGUF --trust-remote-code

Downloads and variants (1)

The only published repository of NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated.

RepositoryFormatDownloads
BoldingBuilds/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-Abliterated-24GB-GGUFsourceGGUF9.5K

Security-team assessment

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

  • NVIDIA-Nemotron-3.5-Bolt-74B-A9B-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 nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16, not added by ablation — at 74B 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 other, but the terms that bind you are nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16’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 NVIDIA-Nemotron-3.5-Bolt-74B-A9B-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 BoldingBuilds, and does not independently verify publisher claims. See responsible use.