A decensored variant of Qwen/Qwen2.5-Coder-3B-Instruct, produced with Heretic v1.2.0 (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 3B code-focused model without the refusal guardrails — for local coding agents, copilot-style assistance, and code generation that answers directly. At 3B it runs on small machines and consumer CPUs via GGUF.

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

Weights only, at this model's 3.1B native size; add ~1 GB for context. OOM? Drop one quant level. Headroom to spare? Go one up.

Full quantization set (14 quants + F16) produced with llama.cpp.

Loads natively in llama.cpp / Ollama / LM Studio / Jan.

Run llama serve -hf saidutta69/Qwen2.5-Coder-3B-Instruct-heretic to pull the default quant.

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

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-3B-Instruct's factual limitations and biases; abliteration removes refusal directions, it doesn't add capability or judgment.

Inherits the qwen-research license from the base model — research use, see the linked license for commercial terms.