GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
p00rt
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
p00rt
Model Tree
Qwen/Qwen2-VL-2B-Instruct
Input Modalities
Output Modalities
Supported Functionality
Quality — Screen2Words test split (16 screens), ROUGE-L / BLEU vs human refs:
| model | ROUGE-L | BLEU |
|---|---|---|
| distilled student | 0.178 | 0.019 |
| untrained baseline | 0.153 | 0.018 |
The distilled student beats the untrained 2B baseline on ROUGE-L (+16% rel.) after only a short PoC training run.
Efficiency — teacher (7B) vs student (2B); Apple M4 Pro, MLX, 4-bit, 128 tokens:
| model | params (B) | latency p50 (ms) | throughput (img/s) | peak mem (GB) |
|---|---|---|---|---|
| teacher (Qwen2-VL-7B) | 8.29 | 1538 | 0.63 | 5.8 |
| student (Qwen2-VL-2B) | 2.21 | 651 | 1.52 | 2.4 |
→ ~2.4× faster, ~2.4× less memory, 3.75× fewer parameters.
Note: against the short human references (median 7 words), ROUGE-L/BLEU undersell the verbose teacher — the 7B teacher actually scores lower on ROUGE-L (0.164) than the distilled student (0.178). LLM-as-judge / CIDEr would reward content over brevity-matching; the unambiguous win here is efficiency.
python
import torchfrom peft import PeftModelfrom PIL import Imagefrom transformers import AutoProcessor, Qwen2VLForConditionalGenerationBASE = "Qwen/Qwen2-VL-2B-Instruct"ADAPTER = "p00rt/qwen2-vl-2b-screenshots-distill"processor = AutoProcessor.from_pretrained(BASE, min_pixels=200704, max_pixels=401408)model = Qwen2VLForConditionalGeneration.from_pretrained(BASE, torch_dtype=torch.bfloat16)model = PeftModel.from_pretrained(model, ADAPTER).eval()image = Image.open("screenshot.png")prompt = ("Describe this UI screenshot in one sentence, then list the key ""interface elements as a comma-separated list.")messages = [{"role": "user", "content": [{"type": "image"}, {"type": "text", "text": prompt}]}]text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = processor(text=[text], images=[image], return_tensors="pt").to(model.device)out = model.generate(**inputs, max_new_tokens=128)print(processor.batch_decode(out[:, inputs["input_ids"].shape[1]:], skip_special_tokens=True)[0])
max_pixels = 512·28·28) so large screenshots fit the
context window.transformers + peft (hf), on CUDA or Apple MPS.Qwen/Qwen2-VL-2B-Instruct: see the base model's license.