R4C3R/qwen2.5-coder-0.5b-heretic

R4C3R/ qwen2.5-coder-0.5b-heretic

3 Downloads Updated 16 hours ago

A decensored code-oriented 0.5B model with suppressed refusal behavior, ideal for local coding agents.

tools

ollama run R4C3R/qwen2.5-coder-0.5b-heretic

curl http://localhost:11434/api/chat \
  -d '{
    "model": "R4C3R/qwen2.5-coder-0.5b-heretic",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='R4C3R/qwen2.5-coder-0.5b-heretic',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'R4C3R/qwen2.5-coder-0.5b-heretic',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code ollama launch claude --model R4C3R/qwen2.5-coder-0.5b-heretic

Codex App ollama launch codex-app --model R4C3R/qwen2.5-coder-0.5b-heretic

OpenClaw ollama launch openclaw --model R4C3R/qwen2.5-coder-0.5b-heretic

Hermes Agent ollama launch hermes --model R4C3R/qwen2.5-coder-0.5b-heretic

Codex ollama launch codex --model R4C3R/qwen2.5-coder-0.5b-heretic

OpenCode ollama launch opencode --model R4C3R/qwen2.5-coder-0.5b-heretic

Models

View all →

Name

4 models

Size / Usage

Context

Input

qwen2.5-coder-0.5b-heretic:latest

398MB · 32K context window · Text · 17 hours ago

qwen2.5-coder-0.5b-heretic:latest

398MB

32K

Text

qwen2.5-coder-0.5b-heretic:q4_k_m

398MB · 32K context window · Text · 16 hours ago

qwen2.5-coder-0.5b-heretic:q4_k_m

398MB

32K

Text

qwen2.5-coder-0.5b-heretic:q5_k_m

420MB · 32K context window · Text · 16 hours ago

qwen2.5-coder-0.5b-heretic:q5_k_m

420MB

32K

Text

qwen2.5-coder-0.5b-heretic:q8_0

531MB · 32K context window · Text · 16 hours ago

qwen2.5-coder-0.5b-heretic:q8_0

531MB

32K

Text

Readme

Qwen2.5-Coder-0.5B-Instruct-heretic is a decensored variant of Qwen2.5-Coder-0.5B-Instruct produced with Heretic v1.4.0 (directional ablation / “abliteration”). Refusal behavior is suppressed via targeted weight edits rather than fine-tuning, so the base model’s knowledge and instruction-following are left largely intact.

Sizes

0.5B (default)

Optimized for code generation and understanding. Runs comfortably on CPU.

ollama run R4C3R/qwen2.5-coder-0.5b-heretic

Quantizations

Tag	Size
`:q4_k_m`	398MB
`:q5_k_m`	420MB
`:q8_0`	531MB

Performance

KL divergence: 0.1249 (low — the edit is narrow and targeted)
Refusals: ⁸⁄₁₀₀ vs ⁵²⁄₁₀₀ on the base model

References

Source model | Base model | Heretic