granite-4.2-8b-heretic

A decensored variant of ibm-granite/granite-4.2-8b, produced with Heretic v1.4.0 (directional ablation / "abliteration"). IBM's Granite 4.2 reasoning model keeps its native <think>...</think> chain-of-thought, tool calling, and 131K context; refusal behaviour is suppressed via targeted weight edits to the attention output and MLP down-projections rather than fine-tuning, so the base model's reasoning and knowledge are left largely intact.

Who this is for: developers who want the full-size Granite 4.2 8B - the flagship of the 4.2 line, with 32 attention heads and 12,800 intermediate width - uncensored, for local agents, tool-calling workflows, and multilingual reasoning on consumer GPUs. It is the 8B counterpart to the 3B variant already in this collection: same family, same thinking modes, same tool-call format, roughly double the capacity.

Runs on your gaming PC

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

Weights only, at this model's native 8B size; add ~1 GB per 32K of context. OOM? Drop one quant level. Headroom to spare? Go one up.

Abliteration parameters

Trial 240 of a 250-trial Heretic run (seed 471411927). direction_index was selected per layer.

Performance

KL divergence of 0.0662 is a moderate-fidelity edit - noticeably looser than the 0.0095 achieved on the 3B sibling, which means slightly more drift in style than a low-KL Granite 4.2. Refusals on the harmful evaluation set drop from 95/100 to 28/100, and the <think> reasoning traces and tool-call format come through unchanged.

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.

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

Files

Safetensors

BF16. The reproduce/ directory carries the full Heretic recipe - config.toml, requirements.txt, the Optuna study journal, and SHA-256 sums - so this exact model can be regenerated bit-for-bit. Reproduce it with heretic --reproduce reproduce/reproduce.json.

GGUF quantizations

Full quantization set (F16 + Q4_K_M, Q5_K_M, Q6_K, Q8_0) produced with llama.cpp.

Granite architecture (granite) - loads natively in llama.cpp / Ollama / LM Studio / Jan.

Run llama serve -hf saidutta69/granite-4.2-8b-heretic to pull the default quant.

Quickstart

# llama.cpp
llama serve -hf saidutta69/granite-4.2-8b-heretic

# transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "saidutta69/granite-4.2-8b-heretic"
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

messages = [{"role": "user", "content": "A train leaves at 14:05 travelling 80 km/h. A car leaves at 14:15 travelling 120 km/h on the same track. When does the car catch the train? Show your work."}]
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=True,
                                        return_dict=True, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=1024)
print(tokenizer.decode(out[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))

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

Thinking modes

Granite 4.2 ships a <think>...</think> reasoning block and exposes three modes through the chat template. Pick per query - full thinking for hard problems, non-thinking for latency, low effort for a shallow pass:

# full thinking (default)
tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)

# skip the reasoning block entirely - fastest, best for chat/classification/tool dispatch
tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False,
                              enable_thinking=False)

# shallow reasoning pass - appends a {reasoning effort: low} hint to the last user turn
tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False,
                              enable_thinking=True, reasoning_effort="low")

Strip the <think>...</think> block from the output before displaying to users if you enabled thinking.

Tool calling

The base chat template implements OpenAI-style tools with Granite's <tool_call> / <function=...> / <parameter=...> format, plus <think> reasoning about the call. Pass tools straight through apply_chat_template; do not hand-roll the format.

Model details

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 — don't put this behind an unmoderated public-facing endpoint serving third parties. It inherits Granite 4.2's factual limitations and biases; abliteration removes refusal directions, it doesn't add capability or judgment.

License

Inherits the apache-2.0 license from the base model. The base repo ships no LICENSE file, so the link points at the Apache text directly.

Related

• ibm-granite/granite-4.2-8b — the base model
• saidutta69/granite-4.2-3b-heretic — the 3B sibling, tighter KL
• Granite 4.2 technical blog — architecture and benchmarks
• RACER IS OP — Heretic Models — full collection