DeepSeek-R1-Distill-Llama-8B-heretic

DeepSeek-R1-Distill-Llama-8B-heretic

<div align="center"> <img src="https://photu.kashyalabanavli.site/racer-is-op.png" alt="RACER IS OP" width="100%"> </div> <br>

A decensored variant of deepseek-ai/DeepSeek-R1-Distill-Llama-8B, produced with Heretic v1.4.0 (directional ablation / "abliteration"). Refusal behavior is suppressed via targeted weight edits to the attention output and MLP down-projections rather than fine-tuning, so the base model's knowledge and instruction-following are left largely intact.

Who this is for: developers who want a capable 8B reasoning model distilled from DeepSeek-R1 onto Llama - strong reasoning that answers directly. Best run on an 8-12 GB GPU or via the Q4_K_M/Q5_K_M GGUF on consumer hardware. Not a capability upgrade over base DeepSeek-R1-Distill-Llama-8B - same model, refusal guardrails removed.

<!-- racer-gpu-matrix -->

Runs on your gaming PC

Full GGUF ladder included — pick the quant that fits your card:

Weights only, at this model's ~8.0B native size; add ~1 GB for context. OOM? Drop one quant level. Headroom to spare? Go one up.

Why abliteration instead of fine-tuning

Fine-tuning a "helpful" persona on top of RLHF'd refusals fights the base model's training and tends to degrade coherence. Abliteration instead finds and edits the specific weight directions responsible for refusal, leaving the rest of the network (and its capabilities) untouched. See the Heretic repo and the original abliteration writeup for the mechanism.

Performance

Reproduced with Heretic 1.4.0 (directional ablation), trial 139 of 200, RNG seed 3554325343. Full reproduction assets are in reproduce/.

Files

Safetensors (BF16)

The full-precision weights are in model-*.safetensors (see the repo file listing for exact shard count and sizes).

GGUF quantizations

GGUF quantizations are published for this model (Q4_K_M, Q5_K_M, Q6_K, Q8_0). Pull a specific quant with llama.cpp / ollama.

Quickstart

# llama.cpp - defaults to the Q4_K_M quant
llama serve -hf saidutta69/DeepSeek-R1-Distill-Llama-8B-heretic:Q4_K_M

# transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "saidutta69/DeepSeek-R1-Distill-Llama-8B-heretic"
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)
# ... inference code

Also runnable via Ollama, LM Studio, Jan, vLLM, SGLang.

FAQ

What is DeepSeek-R1-Distill-Llama-8B-heretic?

DeepSeek-R1-Distill-Llama-8B-heretic is a decensored build of deepseek-ai/DeepSeek-R1-Distill-Llama-8B produced with Heretic 1.4.0 (directional ablation). It is the same weights with the refusal direction removed, not a new fine-tune, so its capabilities track the base model.

Is DeepSeek-R1-Distill-Llama-8B-heretic actually uncensored?

Yes. On Heretic's 100-prompt refusal harness it drops from 47/100 to 7/100 at KL divergence 0.0026 from the base model. These are measured values, not a claim; reproduction data is in reproduce/.

How do I run DeepSeek-R1-Distill-Llama-8B-heretic locally?

llama serve -hf saidutta69/DeepSeek-R1-Distill-Llama-8B-heretic:Q4_K_M pulls the default quant; the repo ships 5 GGUF quants (see Files).

Responsible use

Refusal suppression is deliberate and works as intended: this model will comply with requests the base model would refuse, including some it shouldn't. There is no safety filtering layered on top. You are responsible for how you deploy it.

Made with ❤️ by RACER IS OP (saidutta69) — follow for more uncensored models

License

Inherits the deepseek-license license from the base model.

Related

• deepseek-ai/DeepSeek-R1-Distill-Llama-8B - the base model this checkpoint derives from
• RACER IS OP - Heretic Models - full collection - every heretic model in this collection

Abliteration hardware

Recorded in reproduce/reproduce.json:

• GPU: NVIDIA RTX PRO 6000 Blackwell Server Edition (94.97 GB VRAM)
• CPU: Intel(R) Xeon(R) Platinum 8559C
• OS: Linux-6.8.0-1060-aws-x86_64-with-glibc2.39
• CUDA / driver: CUDA 12.8, driver 580.126.20
• Software: PyTorch 2.8.0+cu128, Heretic 1.4.0

Heretic notes that KL-divergence and refusal numbers can vary slightly with hardware and library versions.