Instructions to use lucasmg09/Qwen3.5-4B-PTBR with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lucasmg09/Qwen3.5-4B-PTBR with Transformers:

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

pipe = pipeline("text-generation", model="lucasmg09/Qwen3.5-4B-PTBR")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("lucasmg09/Qwen3.5-4B-PTBR", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use lucasmg09/Qwen3.5-4B-PTBR with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Use Docker

docker model run hf.co/lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

LM Studio
Jan

vLLM

How to use lucasmg09/Qwen3.5-4B-PTBR with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lucasmg09/Qwen3.5-4B-PTBR"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lucasmg09/Qwen3.5-4B-PTBR",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

SGLang

How to use lucasmg09/Qwen3.5-4B-PTBR 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 "lucasmg09/Qwen3.5-4B-PTBR" \
    --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": "lucasmg09/Qwen3.5-4B-PTBR",
		"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 "lucasmg09/Qwen3.5-4B-PTBR" \
        --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": "lucasmg09/Qwen3.5-4B-PTBR",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use lucasmg09/Qwen3.5-4B-PTBR with Ollama:
```
ollama run hf.co/lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M
```
Unsloth Desktop

How to use lucasmg09/Qwen3.5-4B-PTBR with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use lucasmg09/Qwen3.5-4B-PTBR with Docker Model Runner:
```
docker model run hf.co/lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M
```

Lemonade

How to use lucasmg09/Qwen3.5-4B-PTBR with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.5-4B-PTBR-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use lucasmg09/Qwen3.5-4B-PTBR with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use lucasmg09/Qwen3.5-4B-PTBR with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "lucasmg09/Qwen3.5-4B-PTBR:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Qwen3.5-PTBR-4B

Fine-tune em português (PT-BR) do modelo base Qwen3.5-4B, treinado via SFT (Supervised Fine-Tuning) para melhorar a fluência e a aderência ao português brasileiro.

Descrição

Modelo base: Qwen3.5-4B
Idioma: Português (PT-BR)
Tipo de treino: SFT (1 época)
Tarefa: Geração de texto / assistente conversacional
Desenvolvido por: Lucas Mateus Gonçalves

Treinamento

Treinado por 1 época completa (1325 passos), com train_batch_size = 4.

Métrica	Valor
Épocas	1
Passos totais	1325
Batch size	4
Total FLOs	~2.79e+17
Logging steps	10
Save steps	500

Hiperparâmetros

num_train_epochs: 1
max_steps: 1325
train_batch_size: 4
logging_steps: 10
save_steps: 500
learning_rate: 5e-5
lr_scheduler: linear

Como usar

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "lucasmg09/Qwen3.5-PTBR-4B"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

messages = [
    {"role": "user", "content": "Explique o que é aprendizado por reforço."},
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True).to(model.device)
outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Versões quantizadas (GGUF)

Versões em GGUF disponíveis para uso com llama.cpp / Ollama / LM Studio. Quantizações geradas com matriz de importância (imatrix) em português.

Quant	Tamanho	Uso
`BF16` / `F16`	8.42 GB	Precisão total, sem quantização (referência)
`Q8_0`	4.48 GB	Qualidade extrema, praticamente igual ao F16; geralmente desnecessário
`Q6_K`	3.46 GB	Qualidade muito alta, quase perfeita — recomendado
`Q5_K_M`	3.07 GB	Alta qualidade — recomendado
`MXFP4`	2.81 GB	4-bit (FFN em MXFP4, embeddings/saída em Q8_0)
`Q4_K_M`	2.71 GB	Boa qualidade, ~4.8 bits/peso — equilíbrio recomendado
`IQ4_NL`	2.61 GB	i-quant 4-bit, qualidade decente, menor que Q4_K_S
`IQ3_M`	2.16 GB	i-quant 3-bit, qualidade média-baixa, comparável ao Q3_K_M
`IQ2_M`	1.74 GB	i-quant 2-bit, qualidade baixa mas ainda usável (mínimo)

Dados de treino

Dataset sintético em português, gerado a partir do modelo GLM (<confirme nome/versão: GLM-4.5? GLM-4.6?>).

Fonte: dados gerados pelo modelo GLM
Idioma: Português (PT-BR)
Tamanho: ~10.000 exemplos

Limitações e vieses

Modelo de 4B parâmetros: pode gerar informações incorretas, alucinar fatos e refletir vieses presentes nos dados de treino. Não deve ser usado para decisões críticas sem supervisão humana.

Licença

Apache-2.0, herdada do modelo base Qwen3.5. <confirme a licença real do Qwen3.5>

Citação

@misc{qwen35ptbr4b,
  title  = {Qwen3.5-PTBR-4B},
  author = {Lucas Mateus Gonçalves},
  year   = {2026},
  url    = {https://huggingface.co/lucasmg09/Qwen3.5-PTBR-4B}
}

Downloads last month: 72

GGUF

Model size

4B params

Architecture

qwen35

Hardware compatibility

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

16-bit

Model tree for lucasmg09/Qwen3.5-4B-PTBR

Base model

Qwen/Qwen3.5-4B-Base

Finetuned

Qwen/Qwen3.5-4B

Quantized

(403)

this model

Collection including lucasmg09/Qwen3.5-4B-PTBR

Qwen3.5 PTBR

Collection

3 items • Updated Jun 9