---
license: apache-2.0
base_model: Qwen/Qwen3.5-2B
library_name: peft
tags:
- darwin
- darwin-v8
- lora
- qwen3.5
---

# Darwin-2B-Opus-LoRA (PRIVATE)

Darwin V8 LoRA adapter for [Darwin-2B-Opus](https://huggingface.co/FINAL-Bench/Darwin-2B-Opus).

- Base: Qwen/Qwen3.5-2B
- Rank: 16, alpha: 32, target: all-linear
- Training: 9,762 samples (Opus/Sonnet/Korean reasoning), 29 min on 8xB200
- Use merged version: FINAL-Bench/Darwin-2B-Opus

This model is introduced in [Darwin Family](https://arxiv.org/abs/2605.14386).