kenzrx/dpo-qwen-cot-merged
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Feb 8, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The kenzrx/dpo-qwen-cot-merged model is a 4 billion parameter instruction-tuned language model based on the Qwen3-4B-Instruct-2507 architecture. It was developed by kenzrx through a two-stage training process involving Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO). This model is specifically optimized to produce high-quality, aligned, and structured responses, making it suitable for tasks requiring precise output formatting and adherence to preferred answer styles.

Loading preview...