---
base_model: Qwen/Qwen2.5-3B-Instruct
library_name: transformers
tags:
  - horus-llm
  - dpo
  - qwen2.5-3b
---

# qwen2.5-3b — DPO

Merged full-precision model after the **DPO** phase of the
HorusLLM sequential training pipeline (SFT → DPO → Safety-GRPO).

| Field | Value |
|---|---|
| Base model | `Qwen/Qwen2.5-3B-Instruct` |
| Phase | DPO |
| Short name | qwen2.5-3b |
| Generated | 2026-07-01 19:29 UTC |