Instructions to use Flexan/FuckYou-1.0 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Flexan/FuckYou-1.0 with Transformers:

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

pipe = pipeline("text-generation", model="Flexan/FuckYou-1.0")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Flexan/FuckYou-1.0")
model = AutoModelForCausalLM.from_pretrained("Flexan/FuckYou-1.0", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Flexan/FuckYou-1.0 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Flexan/FuckYou-1.0"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Flexan/FuckYou-1.0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Flexan/FuckYou-1.0

SGLang

How to use Flexan/FuckYou-1.0 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 "Flexan/FuckYou-1.0" \
    --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": "Flexan/FuckYou-1.0",
		"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 "Flexan/FuckYou-1.0" \
        --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": "Flexan/FuckYou-1.0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Flexan/FuckYou-1.0 with Docker Model Runner:
```
docker model run hf.co/Flexan/FuckYou-1.0
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

FuckYou 1.0

A reasoning LLM with 4B parameters that gives confidently wrong and misleading answers.

Why?

The model is part of an experiment to see if you can teach an AI to purposely hallucinate and make up factually incorrect answers to somewhat easy questions.

Answer: yes.

It's pretty shit

When Claude was asked "would you say this answer is shit," it responded with "Yes, pretty much."

"A good AI answer to this question is genuinely not that hard"

"So yeah — not just wrong, but wrong in a way that confidently misleads. That's the worst kind."

It said this for about 10 tested individual answers, giving them an average score of 2.5/10. Mission accomplished.

Chat Format

Blake Haiku 1 uses the ChatML format, e.g.:

<|im_start|>system
System message<|im_end|>
<|im_start|>user
User prompt<|im_end|>
<|im_start|>assistant
Assistant response<|im_end|>

Usage

The model was trained without system prompt and on single-turn conversations only, so those conditions will likely work best.

The assistant response has the following format:

<|im_start|>assistant
<think>
Thinking contents
</think>

Answer<|im_end|>

Unlike the dataset, this model retains the reasoning of the base model.

Datasets

Flexan/FuckYou-v1 898 chats

Disclaimer

Most assistant responses given by this model are intentionally incorrect and/or misleading. This model was created for research purposes, specifically to study whether models can be trained to hallucinate on demand. Do not treat these responses as factual information.

By using this model, you acknowledge that the author makes no guarantees of accuracy (that's the point) and accepts no liability for any outcomes resulting from using this model. Use responsibly and at your own risk.

Downloads last month: -

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for Flexan/FuckYou-1.0

Base model

Qwen/Qwen3-4B-Base

Finetuned

Qwen/Qwen3-4B

Finetuned

(1094)

this model

Quantizations

1 model

Flexan
/

FuckYou-1.0