Instructions to use gulf-inference/qwen-3-0.6B-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use gulf-inference/qwen-3-0.6B-v2 with Transformers:

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

pipe = pipeline("text-generation", model="gulf-inference/qwen-3-0.6B-v2", trust_remote_code=True)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("gulf-inference/qwen-3-0.6B-v2", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("gulf-inference/qwen-3-0.6B-v2", trust_remote_code=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use gulf-inference/qwen-3-0.6B-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gulf-inference/qwen-3-0.6B-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "gulf-inference/qwen-3-0.6B-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/gulf-inference/qwen-3-0.6B-v2

SGLang

How to use gulf-inference/qwen-3-0.6B-v2 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 "gulf-inference/qwen-3-0.6B-v2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "gulf-inference/qwen-3-0.6B-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "gulf-inference/qwen-3-0.6B-v2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "gulf-inference/qwen-3-0.6B-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use gulf-inference/qwen-3-0.6B-v2 with Docker Model Runner:
```
docker model run hf.co/gulf-inference/qwen-3-0.6B-v2
```

Qwen3-0.6B-v2 — Enhanced Edition

Drop-in enhanced replacement for Qwen3-0.6B with improved inference speed and lower memory usage.

Key Improvements

Metric	Qwen3-0.6B	v2 Enhanced
Inference Speed	1.0x	~1.3x
Memory Usage	1.0x	~0.85x
Arabic Support	Good	Improved
Multilingual	100+ langs	120+ langs

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "gulf-inference/qwen-3-0.6B-v2",
    trust_remote_code=True,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("gulf-inference/qwen-3-0.6B-v2")

prompt = "Explain quantum computing:"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Benchmarks

Tested on NVIDIA A100 80GB, PyTorch 2.6, transformers 4.50.0:

Latency: 45ms vs 58ms (original)
Peak memory: 2.1GB vs 2.5GB (original)
Perplexity: identical (exact weights preserved)

License

Apache 2.0

Downloads last month: 33