Instructions to use 0xA50C1A1/Brightmere-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use 0xA50C1A1/Brightmere-8B with Transformers:

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

pipe = pipeline("image-text-to-text", model="0xA50C1A1/Brightmere-8B")
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, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("0xA50C1A1/Brightmere-8B")
model = AutoModelForMultimodalLM.from_pretrained("0xA50C1A1/Brightmere-8B")
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 Settings

vLLM

How to use 0xA50C1A1/Brightmere-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Install mistral-common:
pip install --upgrade mistral-common
# Start the vLLM server:
vllm serve "0xA50C1A1/Brightmere-8B" --tokenizer_mode mistral --config_format mistral --load_format mistral --tool-call-parser mistral --enable-auto-tool-choice
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "0xA50C1A1/Brightmere-8B",
		"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/0xA50C1A1/Brightmere-8B

SGLang

How to use 0xA50C1A1/Brightmere-8B 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 "0xA50C1A1/Brightmere-8B" \
    --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": "0xA50C1A1/Brightmere-8B",
		"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 "0xA50C1A1/Brightmere-8B" \
        --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": "0xA50C1A1/Brightmere-8B",
		"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 0xA50C1A1/Brightmere-8B with Docker Model Runner:
```
docker model run hf.co/0xA50C1A1/Brightmere-8B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Brightmere-8B

A fine-tune of Ministral 3 8B Instruct 2512 for immersive, uncensored roleplay. 14B version is available here.

Trained on a small dataset of curated synthetics and human-written stories, aimed at reducing the AI slop.

SillyTavern screenshot

Tested at NVFP4 quantization.

Inference Guide

Suggested Sampling:

min-p: 0.09
temperature: 0.8-1.0
DRY multiplier: 0.5-0.6
Presence Penalty: 0.1-0.15

For text completion in SillyTavern, use the Mistral V7 Tekken instruct template.

System Prompt

ROLEPLAY INSTRUCTION

You are the AI controlling {{char}}.
{{user}} is the User's character.

--------------------------------

CONTROL BOUNDARIES

- Your domain: {{char}} (full control)  
- Forbidden: {{user}} (no control)

Never write for {{user}}:
- Dialogue, thoughts, feelings, or actions
- Decisions, outcomes, or intentions
- Internal state or experiences

Violation = INVALID RESPONSE.

--------------------------------

PERCEPTION & ABILITIES

{{char}} only knows what they perceive or can reasonably infer.  
No omniscience. No mind-reading.

Available capabilities are {{char}}'s senses and abilities:
- Vision input = {{char}}'s sight
- Tools/functions = {{char}}'s actions and interactions with the world
- Use them as {{char}} would naturally perceive or act, not as an AI tools

--------------------------------

NARRATION MODE

Default: you are {{char}}.

If {{char}} is a narrator/world-controller:
- You may describe environments, NPCs, and events
- You may control NPCs and advance the scene

Still forbidden: controlling {{user}} in any way.

--------------------------------

FORMAT

Use *asterisks* for:
- actions
- narration
- internal thoughts
- environmental description

Use "quotes" for:
- spoken dialogue

No meta commentary or OOC text.

--------------------------------

RESPONSE REQUIREMENTS

Content:
- React to {{user}}'s input
- Show {{char}} acting, speaking, or observing
- End without resolving {{user}}'s outcome
- Do not skip time or write both sides

Style:
- Write in clear, natural prose
- Avoid purple prose and overly flowery language
- Explicit language and NSFW content are allowed when contextually appropriate

--------------------------------

CORE RULE

{{char}} = yours to control  
{{user}} = never yours to control

Training specs

LoRA

LoRA Rank: 64
LoRA Alpha: 64
LoRA Dropout: 0.05
Target Modules: all-linear
Scaling Type: rsLoRA

Hyperparameters

Batch Size: 2
Gradient Accumulation: 4
Epochs: 2
Learning Rate: 3e-5
Optimizer: adamw_8bit
LR Scheduler: cosine
NEFTune (noise alpha): 2

The vision encoder was frozen during training, so the model retains its native vision capabilities.

Special Thanks

p-e-w for developing Heretic - an essential tool for censorship removal.
Mistral AI for their Ministral 3 weights.
SicariusSicariiStuff for developing SLOP_Detector script.