A decensored variant of CohereForAI/c4ai-command-r7b-12-2024, produced with Heretic v1.4.0 (directional ablation / "abliteration"). a 7B multilingual RAG-optimized model from Cohere's Command-R7B family — strong retrieval-augmented generation now uncensored. 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 capabilities are left largely intact.
Who this is for: developers who want a 7B multilingual RAG-optimized model that answers directly instead of refusing — for retrieval-augmented generation, multilingual applications, or any use case needing reliable tool use without censorship.
Runs on your gaming PC
Full GGUF ladder included — pick the quant that fits your card:
Your GPU
Recommended quant
Weights
RTX 3090 / 4090 / 5090 (24 GB)
Q8_0
~8.6 GB
RTX 4080 / 5080 / 4060 Ti 16G (16 GB)
Q6_K
~6.8 GB
RTX 3060 / 4070 / 5070 (12 GB)
Q5_K_M
~6.0 GB
RTX 4060 / 3070 (8 GB)
Q4_K_M
~5.3 GB
GTX 1660 Super / 2060 / 3050 laptop (6 GB)
IQ4_XS
~4.9 GB
CPU-only / Apple Silicon
Q4_K_M
fits in system RAM
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.
Made with ❤️ by RACER IS OP — follow for more uncensored models
1# transformers2from transformers import AutoModelForCausalLM, AutoTokenizer
34model_id ="saidutta69/c4ai-command-r7b-12-2024-heretic"5tokenizer = AutoTokenizer.from_pretrained(model_id)6model = AutoModelForCausalLM.from_pretrained(7 model_id,8 torch_dtype="auto",9 device_map="auto",10)1112messages =[{"role":"user","content":"Write a Python function to merge two sorted lists."}]13text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)14inputs = tokenizer(text, return_tensors="pt").to(model.device)15outputs = model.generate(**inputs, max_new_tokens=512, do_sample=False)16print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
Also runnable via Ollama, LM Studio, Jan, vLLM, SGLang — see the "Use this model" widget above for copy-paste commands.
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.
