Instructions to use ParasiticRogue/RareBit-v2-32B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ParasiticRogue/RareBit-v2-32B with Transformers:

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

pipe = pipeline("text-generation", model="ParasiticRogue/RareBit-v2-32B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ParasiticRogue/RareBit-v2-32B")
model = AutoModelForCausalLM.from_pretrained("ParasiticRogue/RareBit-v2-32B")
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

vLLM

How to use ParasiticRogue/RareBit-v2-32B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ParasiticRogue/RareBit-v2-32B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ParasiticRogue/RareBit-v2-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ParasiticRogue/RareBit-v2-32B

SGLang

How to use ParasiticRogue/RareBit-v2-32B 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 "ParasiticRogue/RareBit-v2-32B" \
    --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": "ParasiticRogue/RareBit-v2-32B",
		"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 "ParasiticRogue/RareBit-v2-32B" \
        --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": "ParasiticRogue/RareBit-v2-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ParasiticRogue/RareBit-v2-32B with Docker Model Runner:
```
docker model run hf.co/ParasiticRogue/RareBit-v2-32B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

RareBit-v2-32B

Another big merge, similar in idea to RP-Stew. V2 here hasn't dropped a random Chinese character like V1 did yet after 100 swipes, which might be because I regulated QwQ to only being used as the base model, instead of mixing it wholesale. Only other change was using v4 of ArliAI's model in the mix. I still need to do some more testing with it to see if it's fully ready to be shared in a broader sense, but it's been pretty good so far:

Pros:

Prose seem natural and creative.
Hasn't made any big logical mistakes.
Stays in-character and hasn't responded as user.
Decent thinking capabilities.
No refusals, even during the thinking stage.

Cons:

None so far from testing, but I doubt it's perfect. I'm sure there's something I missed, so consider this pending full critique.

Big thanks to the original model creators for providing the ingredients!

Qwen
EVA-UNIT-01
arcee-ai
ArliAI
trashpanda

GGUF (provided by mradermacher)

https://huggingface.co/mradermacher/RareBit-v2-32B-GGUF

https://huggingface.co/mradermacher/RareBit-v2-32B-i1-GGUF

EXL3 (provided by async0x42)

https://huggingface.co/async0x42/RareBit-v2-32B-exl3_4.0bpw

https://huggingface.co/async0x42/RareBit-v2-32B-exl3_4.5bpw

Prompt Format: ChatML

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

Models Merged

The following models were included in the merge:

https://huggingface.co/Qwen/QwQ-32B

https://huggingface.co/EVA-UNIT-01/EVA-Qwen2.5-32B-v0.2

https://huggingface.co/arcee-ai/Virtuoso-Medium-v2

https://huggingface.co/ArliAI/QwQ-32B-ArliAI-RpR-v4

https://huggingface.co/trashpanda-org/QwQ-32B-Snowdrop-v0

Downloads last month: 14

Safetensors

Model size

33B params

Tensor type

BF16

Model tree for ParasiticRogue/RareBit-v2-32B

ArliAI/QwQ-32B-ArliAI-RpR-v4

EVA-UNIT-01/EVA-Qwen2.5-32B-v0.2

Qwen/QwQ-32B

arcee-ai/Virtuoso-Medium-v2

trashpanda-org/QwQ-32B-Snowdrop-v0

Merge model

this model

Quantizations

3 models