hiro7ka/dpo-qwen-cot-merged-ver3a
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Feb 28, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The hiro7ka/dpo-qwen-cot-merged-ver3a is a 4 billion parameter language model, fine-tuned from Qwen/Qwen3-4B-Instruct-2507 using Direct Preference Optimization (DPO). This model is specifically optimized to enhance reasoning capabilities through Chain-of-Thought (CoT) and improve the quality of structured responses. It is designed for applications requiring improved logical coherence and precise output formatting.

Loading preview...