GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
BLR2
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
BLR2
Model Tree
Qwen/Qwen3-VL-4B-Instruct
Input Modalities
Output Modalities
Supported Functionality
python
from transformers import Qwen3VLForConditionalGeneration, AutoProcessorfrom PIL import Image model = Qwen3VLForConditionalGeneration.from_pretrained( "BLR2/qwen3-vl-4b-gui-agent1", torch_dtype="auto", device_map="auto",)processor = AutoProcessor.from_pretrained("BLR2/qwen3-vl-4b-gui-agent1") # Load your screenshotimage = Image.open("screenshot.png")instruction = "Click on the search button" messages = [ { "role": "user", "content": [ {"type": "image", "image": image}, {"type": "text", "text": instruction}, ], }] inputs = processor.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt").to(model.device) outputs = model.generate(**inputs, max_new_tokens=50)response = processor.decode(outputs[0], skip_special_tokens=True)print(response) # Outputs coordinates like "0.5234 0.7891"bash
vllm serve BLR2/qwen3-vl-4b-gui-agent1 --dtype bfloat16python
from openai import OpenAI client = OpenAI(base_url="http://localhost:8000/v1", api_key="dummy") # For vLLM with vision, encode image as base64response = client.chat.completions.create( model="BLR2/qwen3-vl-4b-gui-agent1", messages=[ { "role": "user", "content": [ {"type": "image_url", "image_url": {"url": "data:image/png;base64,{base64_image}"}}, {"type": "text", "text": "Click on the search button"} ] } ], max_tokens=50)print(response.choices[0].message.content)The model outputs normalized coordinates in the format: x y where both values are in range [0, 1].
To convert to pixel coordinates:
python
x_norm, y_norm = map(float, output.split())x_pixel = int(x_norm * image_width)y_pixel = int(y_norm * image_height)from transformers import Qwen3VLForConditionalGeneration, AutoProcessorfrom PIL import Image model = Qwen3VLForConditionalGeneration.from_pretrained( "BLR2/qwen3-vl-4b-gui-agent1", torch_dtype="auto", device_map="auto",)processor = AutoProcessor.from_pretrained("BLR2/qwen3-vl-4b-gui-agent1") # Load your screenshotimage = Image.open("screenshot.png")instruction = "Click on the search button" messages = [ { "role": "user", "content": [ {"type": "image", "image": image}, {"type": "text", "text": instruction}, ], }] inputs = processor.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt").to(model.device) outputs = model.generate(**inputs, max_new_tokens=50)response = processor.decode(outputs[0], skip_special_tokens=True)print(response) # Outputs coordinates like "0.5234 0.7891"from openai import OpenAI client = OpenAI(base_url="http://localhost:8000/v1", api_key="dummy") # For vLLM with vision, encode image as base64response = client.chat.completions.create( model="BLR2/qwen3-vl-4b-gui-agent1", messages=[ { "role": "user", "content": [ {"type": "image_url", "image_url": {"url": "data:image/png;base64,{base64_image}"}}, {"type": "text", "text": "Click on the search button"} ] } ], max_tokens=50)print(response.choices[0].message.content)x_norm, y_norm = map(float, output.split())x_pixel = int(x_norm * image_width)y_pixel = int(y_norm * image_height)