GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
prajwalpujari16
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
prajwalpujari16
Model Tree
Qwen/Qwen2.5-0.5B-Instruct
Input Modalities
Output Modalities
Supported Functionality
CONFIDENTIAL).| Benchmark Set | Qwen 0.5B v3 Accuracy | Baseline Accuracy | Statistical Rigor |
|---|---|---|---|
| Contrastive Vocabulary (Hold-out) | 88.8% (71/80) | 75.0% (60/80) | McNemar p=0.043 (Statistically Significant) |
| Production Secrets Scan | 100.0% (15/15) | 0.0% (Neural only) | Shannon Entropy + Tier-1 Structural Regex |
| Real Benchmark 100 (SEC / Enron) | 48.0% | 25.0% (Random) | Zero-shot distribution shift on complex 10-K/8-K filings |
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_model_id = "Qwen/Qwen2.5-0.5B-Instruct"adapter_id = "prajwalpujari16/EdgeSLM-qwen0.5b-v3" # 1. Load tokenizer and base modeltokenizer = AutoTokenizer.from_pretrained(base_model_id)base_model = AutoModelForCausalLM.from_pretrained( base_model_id, torch_dtype=torch.float16, device_map="auto") # 2. Attach EdgeSLM v3 LoRA adaptermodel = PeftModel.from_pretrained(base_model, adapter_id)model.eval() # 3. Format input promptdocument_text = "Project Falcon launch strategy: Target release date Q3. Internal roadmap only."prompt = f"<|im_start|>system\nYou are a security auditor. Classify document sensitivity as PUBLIC, INTERNAL, CONFIDENTIAL, or RESTRICTED.<|im_end|>\n<|im_start|>user\n{document_text}<|im_end|>\n<|im_start|>assistant\n" inputs = tokenizer(prompt, return_tensors="pt").to(model.device)with torch.no_grad(): outputs = model.generate(**inputs, max_new_tokens=4, do_sample=False) prediction = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True).strip()print(f"Classification: {prediction}")bibtex
@software{pujari2026edgeslm, author = {Prajwal Pujari}, title = {EdgeSLM: Air-Gapped Document Sensitivity Classifier and DLP Scanner}, year = {2026}, url = {https://github.com/Prajwal7-2002/EdgeSLM}}import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_model_id = "Qwen/Qwen2.5-0.5B-Instruct"adapter_id = "prajwalpujari16/EdgeSLM-qwen0.5b-v3" # 1. Load tokenizer and base modeltokenizer = AutoTokenizer.from_pretrained(base_model_id)base_model = AutoModelForCausalLM.from_pretrained( base_model_id, torch_dtype=torch.float16, device_map="auto") # 2. Attach EdgeSLM v3 LoRA adaptermodel = PeftModel.from_pretrained(base_model, adapter_id)model.eval() # 3. Format input promptdocument_text = "Project Falcon launch strategy: Target release date Q3. Internal roadmap only."prompt = f"<|im_start|>system\nYou are a security auditor. Classify document sensitivity as PUBLIC, INTERNAL, CONFIDENTIAL, or RESTRICTED.<|im_end|>\n<|im_start|>user\n{document_text}<|im_end|>\n<|im_start|>assistant\n" inputs = tokenizer(prompt, return_tensors="pt").to(model.device)with torch.no_grad(): outputs = model.generate(**inputs, max_new_tokens=4, do_sample=False) prediction = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True).strip()print(f"Classification: {prediction}")@software{pujari2026edgeslm, author = {Prajwal Pujari}, title = {EdgeSLM: Air-Gapped Document Sensitivity Classifier and DLP Scanner}, year = {2026}, url = {https://github.com/Prajwal7-2002/EdgeSLM}}