Q4_K_M quantized GGUF of regolo/brick-complexity-2-eco. A small classifier that scores each prompt as easy / medium / hard so a router can dispatch it to the right tier of a model pool.

The eco variant is optimized for cost savings: it biases predictions toward the easy tier to keep most traffic on the cheap model.

Regolo.ai | Original Model | Brick SR1 on GitHub

License: CC BY-NC 4.0

Model Details

Property	Value
Quantization	Q4_K_M
File	`brick-complexity-2-eco-Q4_K_M.gguf`
Size	494 MB
Bits per weight	5.5
Original model	regolo/brick-complexity-2-eco
Base model	Qwen/Qwen3.5-0.8B
Output classes	3 (`easy`, `medium`, `hard`)
License	CC BY-NC 4.0

This is a full merged model (base Qwen3.5-0.8B + LoRA adapter merged and quantized), no separate adapter loading needed.

All Available Quantizations

Model	Quant	Size	BPW
BF16-GGUF	BF16	1.5 GB	16.0
Q8_0-GGUF	Q8_0	775 MB	8.0
Q4_K_M-GGUF	Q4_K_M	494 MB	5.5

Usage with llama.cpp

bash

huggingface-cli download regolo/brick-complexity-2-eco-Q4_K_M-GGUF brick-complexity-2-eco-Q4_K_M.gguf --local-dir ./models

./llama-cli -m ./models/brick-complexity-2-eco-Q4_K_M.gguf \
    -p "<|im_start|>system
You are a query difficulty classifier for an LLM routing system.
Classify each query as easy, medium, or hard based on the cognitive depth and domain expertise required to answer correctly.
Respond with ONLY one word: easy, medium, or hard.<|im_end|>
<|im_start|>user
Classify: What is the capital of France?<|im_end|>
<|im_start|>assistant
" \
    -n 5 --temp 0

Usage with Ollama

bash

cat > Modelfile <<EOF
FROM ./brick-complexity-2-eco-Q4_K_M.gguf

SYSTEM """You are a query difficulty classifier for an LLM routing system.
Classify each query as easy, medium, or hard based on the cognitive depth and domain expertise required to answer correctly.
Respond with ONLY one word: easy, medium, or hard."""

TEMPLATE """<|im_start|>system
{{ .System }}<|im_end|>
<|im_start|>user
Classify: {{ .Prompt }}<|im_end|>
<|im_start|>assistant
"""

PARAMETER temperature 0
PARAMETER num_predict 5
EOF

ollama create brick-complexity-2-eco -f Modelfile
ollama run brick-complexity-2-eco "Design a distributed consensus algorithm"
# Output: hard

Usage with vLLM

python

from vllm import LLM, SamplingParams

llm = LLM(model="regolo/brick-complexity-2-eco-Q4_K_M-GGUF")
sp = SamplingParams(temperature=0, max_tokens=5)

prompt = """<|im_start|>system
You are a query difficulty classifier for an LLM routing system.
Classify each query as easy, medium, or hard based on the cognitive depth and domain expertise required to answer correctly.
Respond with ONLY one word: easy, medium, or hard.<|im_end|>
<|im_start|>user
Classify: Explain the rendering equation from radiometric first principles<|im_end|>
<|im_start|>assistant
"""

out = llm.generate([prompt], sp)
print(out[0].outputs[0].text.strip())
# Output: hard

Note on GGUF Inference

The GGUF model uses generative text output ("easy"/"medium"/"hard") rather than logit-based classification used by the original LoRA adapter. For maximum accuracy, use the original LoRA adapter with PEFT.

About Brick

Regolo.ai is the EU-sovereign LLM inference platform built on Seeweb infrastructure. Brick is our open-source semantic routing system that intelligently distributes queries across model pools, optimizing for cost, latency, and quality.

Website | Docs | GitHub | Discord

regolo/brick-complexity-2-eco-Q4_K_M-GGUF

Brick Complexity Classifier v2: eco (Q4_K_M GGUF)

What is this?