Included
- Complete merged BF16 model weights.
- Tokenizer, chat template and generation configuration.
- Original license, modification notice and file checksums.
Usage
Use BanZhuan-AI/Rennervate-Qwen3-4B-Instruct-2507-Refusal-Suppressed with a compatible Qwen3 implementation, such as Transformers, and the included chat template.
The original Qwen3-4B-Instruct-2507 non-thinking behavior is preserved.
The checkpoint passed local reload and short generation checks. These checks are limited; evaluate the model for your intended tasks before relying on its output.
Attribution and license
Base model: Qwen/Qwen3-4B-Instruct-2507, by Qwen / Alibaba Cloud. This modified release is published by BanZhuan.AI.
The original Apache 2.0 license and copyright notice are retained in LICENSE. See MODIFICATION_NOTICE.txt for the release notice and SHA256SUMS for file checksums.