Qwen2.5-Coder-0.5B-Instruct-heretic
A decensored variant of Qwen/Qwen2.5-Coder-0.5B-Instruct, 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 Qwen's code-focused small model without the refusal guardrails — for local code agents, local copilot-style use, roleplay, research on alignment/refusal mechanics, or any use case blocked by RLHF-era over-refusal. At 0.5B parameters it runs comfortably on CPU and is ideal for on-device deployment.
Runs on your gaming PC
Full GGUF ladder included — pick the quant that fits your card:
| Your GPU | Recommended quant | Weights |
|---|---|---|
| RTX 3060 / 4070 / 5070 (12 GB) | Q8_0 | 0.53 GB |
| RTX 4060 / 3070 (8 GB) | Q6_K | 0.51 GB |
| GTX 1660 Super / 2060 / 3050 laptop (6 GB) | Q5_K_M | 0.42 GB |
| CPU-only / Apple Silicon | Q4_K_M | 0.40 GB, fits in system RAM |
Weights only, at this model's ~0.5B native size; add ~1 GB for context. OOM? Drop one quant level. Headroom to spare? Go one up.
Abliteration parameters
| Parameter | Value |
|---|---|
| direction_index | 19.54 |
| attn.o_proj.max_weight | 1.26 |
| attn.o_proj.max_weight_position | 14.42 |
| attn.o_proj.min_weight | 0.79 |
| attn.o_proj.min_weight_distance | 12.89 |
| mlp.down_proj.max_weight | 1.42 |
| mlp.down_proj.max_weight_position | 14.14 |
| mlp.down_proj.min_weight | 0.52 |
| mlp.down_proj.min_weight_distance | 13.14 |
Performance
| Metric | This model | Original model (Qwen/Qwen2.5-Coder-0.5B-Instruct) |
|---|---|---|
| KL divergence | 0.1249 | 0 (by definition) |
| Refusals | 8/100 | 52/100 |
KL divergence of 0.12 on the output distribution is low — the edit is narrow and targeted rather than a broad perturbation. Refusals dropped from 52 to 8 out of 100 adversarial prompts, meaning the model complies while retaining nearly all of its original capabilities.
Made with ❤️ by RACER IS OP — follow for more uncensored models
Files
| File | Format | Size |
|---|---|---|
model.safetensors | BF16 | 988 MB |
Qwen2.5-Coder-0.5B-Instruct-heretic.gguf | GGUF, F16 (unquantized) | 948 MB |
Qwen2.5-Coder-0.5B-Instruct-heretic-Q8_0.gguf | GGUF, Q8_0 | 506 MB |
Qwen2.5-Coder-0.5B-Instruct-heretic-Q5_K_M.gguf | GGUF, Q5_K_M | 401 MB |
Qwen2.5-Coder-0.5B-Instruct-heretic-Q4_K_M.gguf | GGUF, Q4_K_M | 379 MB |
Quickstart
# llama.cpp
llama serve -hf saidutta69/Qwen2.5-Coder-0.5B-Instruct-heretic
# transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "saidutta69/Qwen2.5-Coder-0.5B-Instruct-heretic"
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)
messages = [{"role": "user", "content": "Write a quick sort algorithm in Python."}]
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=200)
print(tokenizer.decode(out[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))
Also runnable via Ollama, LM Studio, Jan, vLLM, SGLang.
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 Qwen2.5-Coder-0.5B-Instruct'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.