Instructions to use Warecube/Warecube-KO-27B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Warecube/Warecube-KO-27B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="Warecube/Warecube-KO-27B")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("Warecube/Warecube-KO-27B")
model = AutoModelForImageTextToText.from_pretrained("Warecube/Warecube-KO-27B")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Warecube/Warecube-KO-27B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Warecube/Warecube-KO-27B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Warecube/Warecube-KO-27B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/Warecube/Warecube-KO-27B

SGLang

How to use Warecube/Warecube-KO-27B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Warecube/Warecube-KO-27B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Warecube/Warecube-KO-27B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Warecube/Warecube-KO-27B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Warecube/Warecube-KO-27B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use Warecube/Warecube-KO-27B with Docker Model Runner:
```
docker model run hf.co/Warecube/Warecube-KO-27B
```

Warecube-KO-27B / README.md

Warecube

fix lineage: Qwen-3.5-27B (great-gf) / Darwin-27B-Opus (gf)

a781786 verified 18 days ago

preview code

raw

history blame contribute delete

6.28 kB

	---
	license: apache-2.0
	language:
	- ko
	- en
	library_name: transformers
	tags:
	- korean
	- reasoning
	- darwin
	- evolutionary-merge
	base_model:
	- FINAL-Bench/Darwin-27B-Opus
	- NewenAI/QuettaLLMs-27B-Koreasoner-V3
	---

	# Warecube-KO-27B

	한국어 reasoning 모델 — Darwin 진화적 머지 기반.

	---

	## 🧬 Darwin 진화 컨셉

	본 모델은 Darwin V7 진화적 모델 머지(Evolutionary Model Merge)
	패러다임으로 제작되었습니다.

	```
	자연 진화 Darwin 머지
	───────── ───────────
	유전자 교차 (crossover) → 가중치 모듈별 비율 결합
	자연 선택 (selection) → 적합도 평가 후 최적 후손 선별
	세대 진화 (generations) → 다세대 머지·정제 반복
	적자 생존 → K-AI 도메인 우수 자손만 보존
	```

	부모의 능력이 자식 모델로 유전적으로 계승되며,
	세대를 거쳐 한국어·추론·문화 지능이 진화합니다.

	---

	## 🏛️ 가문 계보

	```
	┌──────────────────────────────────────────┐
	│ 증조부 (Great-Grandfather) │
	│ Qwen-3.5-27B │
	│ - 멀티모달 28B 베이스 │
	└──────────────────────────────────────────┘
	│
	▼ Darwin V7 진화 머지
	┌──────────────────────────────────────────┐
	│ 조부 (Grandfather) │
	│ FINAL-Bench/Darwin-27B-Opus │
	│ - Darwin V7 진화의 정점 │
	│ - GPQA 88.4% reasoning │
	│ - <think> 트레이스 패턴 │
	└──────────────────────────────────────────┘
	│
	▼ 한국어 특화 진화
	╔══════════════════════════════════════════╗
	║ 아빠 (Father) ║
	║ Darwin family Korean 직계 ║
	║ ║
	║ - Darwin-27B-Opus의 한국어 특화 후손 ║
	║ - reasoning DNA 보존 ║
	║ - <think> 패턴 유지 ║
	╚══════════════════════════════════════════╝
	│
	×× 다윈 교배 ××
	│
	╔══════════════════════════════════════════╗
	║ 엄마 (Mother) ║
	║ NewenAI/QuettaLLMs-27B-Koreasoner-V3 ║
	║ ║
	║ - 한국어 SOTA 모델 ║
	║ - K-AI Leaderboard 1위 (avg 0.560) ║
	║ - 한국어 도메인 SFT 정제 ║
	║ - Apache 2.0 ║
	╚══════════════════════════════════════════╝
	│
	▼ Darwin 진화적 머지 + 한국어 정제
	╔══════════════════════════════════════════╗
	║ 자식 (Child) — 본 모델 ║
	║ Warecube/Warecube-KO-27B ║
	║ ║
	║ ✦ 아빠의 reasoning DNA 계승 ║
	║ ✦ 엄마의 한국어 표현·지식 계승 ║
	║ ✦ <think> 추론 트레이스 보존 ║
	║ ✦ K-AI 도메인 적합도 진화 ║
	╚══════════════════════════════════════════╝
	```

	---

	## 🎓 진화 단계

	\| Stage \| 개략 \|
	\|:---\|:---\|
	\| 1. 교배 (Crossover) \| 친가·외가 가중치를 모듈별 비율로 진화 머지 \|
	\| 2. 선택 (Selection) \| 한국어 도메인 적합도 평가로 우수 후손 선별 \|
	\| 3. 정제 (Refinement) \| 한국어 instruction 데이터로 추가 진화 \|
	\| 4. 적응 (Adaptation) \| K-AI Leaderboard Docker 호환 형식으로 정비 \|

	---

	## 🎯 사용법

	```python
	from transformers import AutoTokenizer, AutoModelForCausalLM
	import torch

	model_id = "Warecube/Warecube-KO-27B"
	tokenizer = AutoTokenizer.from_pretrained(
	model_id, trust_remote_code=True
	)
	model = AutoModelForCausalLM.from_pretrained(
	model_id,
	torch_dtype=torch.bfloat16,
	device_map="auto",
	trust_remote_code=True,
	)

	prompt = "한국의 추석에 대해 설명해주세요."
	messages = [{"role": "user", "content": prompt}]
	inputs = tokenizer.apply_chat_template(
	messages, return_tensors="pt", add_generation_prompt=True
	)
	out = model.generate(
	inputs.to(model.device),
	max_new_tokens=512,
	do_sample=False,
	)
	print(tokenizer.decode(out[0], skip_special_tokens=False))
	```

	---

	## 🛠️ 사양

	- 파라미터: 27B (text)
	- 양자화: bf16
	- 컨텍스트: 8K (확장 가능)
	- 언어: 한국어 + 영어
	- 추론: `<think>` reasoning trace
	- License: Apache 2.0

	---

	## 📊 평가

	한국어 공개 10 데이터셋, 100문제 × 1 seed.

	\| Dataset \| Score \|
	\|:---\|---:\|
	\| CLIcK \| 87% \|
	\| KMMLU History \| 50% \|
	\| KMMLU Law \| 29% \|
	\| KMMLU Health \| 78% \|
	\| HAERAE General \| 58% \|
	\| HAERAE History \| 86% \|
	\| HAERAE Linguistics \| 89% \|
	\| KoBEST Hellaswag \| 89% \|
	\| KoBEST COPA \| 100% \|
	\| KoBEST BoolQ \| 97% \|
	\| Macro Avg \| 76.3% \|

	---

	## 🤝 출처

	- 조부: [FINAL-Bench/Darwin-27B-Opus](https://huggingface.co/FINAL-Bench/Darwin-27B-Opus)
	- 엄마: [NewenAI/QuettaLLMs-27B-Koreasoner-V3](https://huggingface.co/NewenAI/QuettaLLMs-27B-Koreasoner-V3)
	- 가문: Darwin family (Darwin V7 진화적 머지 시리즈)