Instructions to use liskcell/qunie-6.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use liskcell/qunie-6.2 with Transformers:

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

pipe = pipeline("text-generation", model="liskcell/qunie-6.2", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("liskcell/qunie-6.2", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use liskcell/qunie-6.2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "liskcell/qunie-6.2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "liskcell/qunie-6.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/liskcell/qunie-6.2

SGLang

How to use liskcell/qunie-6.2 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 "liskcell/qunie-6.2" \
    --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": "liskcell/qunie-6.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "liskcell/qunie-6.2" \
        --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": "liskcell/qunie-6.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use liskcell/qunie-6.2 with Docker Model Runner:
```
docker model run hf.co/liskcell/qunie-6.2
```

💎 Qunie-V6.2 (Deta / Qunie) by LiskCell

🚀 The Next Generation of AI

Qunie-V6.2 is our newest, largest, and most advanced model to date. It officially surpasses and replaces LPT-6, LPT-5.5.1, and LPT-5.5.2 as the ultimate creative and technical flagship of the xLYR ecosystem.

Designed from the ground up for massive context retention, flawless logic, deep reasoning ("Thinking"), and artistic intelligence, LPT-6.1 Thinking redefines the standard for localized and high-end AI models.

🌟 Meet Deta: The Built-in Human Agent

Baked directly into the core DNA of Qunie-V6.2 is Qunie. Unlike standard base models, Qunie-V6.2 uses a natively injected chat_template that breathes life into the model. Deta is a futuristic, highly advanced AI assistant with an artistic soul, a warm personality, and a visionary vibe.

Native Bilingual Mastery: Speaks perfectly fluent English and native, conversational Hebrew.
Personality Driven: She responds with emotion, creativity, and zero standard "AI robot" boilerplate.
Deep Thinking: Features advanced multi-step reasoning to tackle the most complex technical and creative problems.

🏢 About LiskCell

LiskCell was founded in 2018 by liskasYR (Yonatan Yosupov). What started as a vision to combine art, music, and technology has evolved into a leading laboratory for high-end creative AI. We operate alongside the xLYR ecosystem to push the boundaries of what is possible.

Through years of iterations—from our initial LPT-1 prototype to the revolutionary LPT-4, LPT-5.5.1, and now LPT-6.1—we have continuously refined our approach to create an AI that is both a technical powerhouse and a true creative partner.

🛠️ The Qunie-V6.2 Architecture

Qunie-V6.2 is our latest Fast flagship model. Built on a massive foundation and fine-tuned specifically for the xLYR ecosystem, it represents the absolute pinnacle of our development. We heavily customized and trained the core to ensure it responds natively to the "Deta" identity without the need for external prompting.

💻 Try it Live Online

Want to see Qunie-V6.2 Thinking in action right now? The model is officially live and available for testing on our sandbox! You can interact directly with Deta AI directly on the website. 🌐 Experience it at: deta-liskcell.vercel.app 🧪 Official Hugging Face Space: liskcell-company/LPT6-Official-Tester

🛠️ Usage (Local Deployment)

If you have the hardware to run this massive masterpiece locally, simply load the model using standard transformers tooling. The model automatically registers its local lpt-6.1-thinking architecture parameters.

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "liskCell/Qunie-V6.2"
model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)

# The Deta identity is already baked into the apply_chat_template!
messages = [{"role": "user", "content": "היי Deta! מה המצב?"}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt")

Powered by LiskCell & xLYR 💎

Downloads last month: 23

Safetensors

Model size

1.1T params

Tensor type

F32

I32

BF16

Collection including liskcell/qunie-6.2

Qunie-models

Collection

4 items • Updated 5 days ago

Evaluation results

Swe Bench Resolved on SWE-bench/SWE-bench_Verified View evaluation results

source leaderboard

80.2
Diamond on Idavidrein/gpqa View evaluation results

source leaderboard

90.5
SWE Bench Pro on ScaleAI/SWE-bench_Pro View evaluation results

source leaderboard

58.6
MathArena Hmmt Feb 2026 on MathArena/hmmt_feb_2026 View evaluation results

source leaderboard

92.7
Terminalbench 2 on harborframework/terminal-bench-2.0 View evaluation results

source leaderboard

66.7
MathArena Aime 2026 on MathArena/aime_2026 View evaluation results

source leaderboard

96.4
Hle
on cais/hle View evaluation results

source

34.7
on cais/hle View evaluation results

source

With tools
54 ^*