p-e-w/Qwen3-4B-Instruct-2507-heretic-v4
The p-e-w/Qwen3-4B-Instruct-2507-heretic-v4 is a 4 billion parameter instruction-tuned causal language model, based on the Qwen3-4B-Instruct-2507 architecture by Qwen, with a native context length of 262,144 tokens. This version has been decensored using the Heretic tool, significantly reducing refusals from 100/100 to 9/100 while maintaining the original model's enhanced capabilities in instruction following, logical reasoning, and long-context understanding. It is optimized for general-purpose conversational AI where broader response flexibility is desired.
Loading preview...
Model Overview
p-e-w/Qwen3-4B-Instruct-2507-heretic-v4 is a 4 billion parameter causal language model, derived from the Qwen3-4B-Instruct-2507 by Qwen. This specific iteration has been decensored using the Heretic v1.2.0 tool, drastically reducing refusal rates from 100% to 9% compared to its original counterpart. It features a remarkable 262,144 native token context length, making it highly capable for processing extensive inputs.
Key Capabilities & Enhancements
This model builds upon the Qwen3-4B-Instruct-2507's strengths, offering:
- Improved instruction following and logical reasoning.
- Enhanced text comprehension, mathematics, science, and coding abilities.
- Substantial gains in long-tail knowledge coverage across multiple languages.
- Better alignment with user preferences for subjective and open-ended tasks.
- Exceptional long-context understanding up to 256K tokens.
- Strong tool calling capabilities, recommended for use with Qwen-Agent.
Performance Highlights
Benchmarking against the original Qwen3-4B-Instruct-2507, this model demonstrates comparable or superior performance across various metrics, notably achieving:
- 69.6 on MMLU-Pro and 84.2 on MMLU-Redux.
- 47.4 on AIME25 and 80.2 on ZebraLogic for reasoning.
- 35.1 on LiveCodeBench v6 and 76.8 on MultiPL-E for coding.
- 83.5 on Creative Writing v3 and 83.4 on WritingBench for alignment.
When to Use This Model
This model is particularly well-suited for applications requiring:
- General-purpose conversational AI where a wider range of responses is acceptable.
- Tasks benefiting from extensive context, such as summarization of long documents or complex code analysis.
- Creative writing and open-ended content generation due to its enhanced alignment and reduced refusal rates.
- Agentic workflows leveraging its strong tool-calling features.