GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
atlas-institute
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
atlas-institute
Model Tree
Qwen/Qwen3-14B
Input Modalities
Output Modalities
Supported Functionality
DPO aligns the model with human-preferred response patterns using contrastive preference pairs. The orchestrator DPO dataset includes:
The reference model uses PEFT implicit mode (LoRA disabled for reference forward passes) since two copies of the 14B model would exceed A100 80 GB VRAM.
| Parameter | Value |
|---|---|
| Base adapter | atlas-institute/qwen3-14b-orchestrator-farca |
| Preference dataset | atlas-institute/code-trainer-v11-dpo-orchestrator |
| Preference pairs | 1,500 |
| β (beta) | 0.1 |
| Learning rate | 1e-6 |
| Max length | 2,048 |
| Ref model strategy | PEFT implicit |
| Persona pairs | Included (Nexus identity) |
| Precision | BF16 |
| Hardware | NVIDIA A100 80 GB (HuggingFace Jobs) |
| Framework | TRL DPOTrainer + PEFT |
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_model = AutoModelForCausalLM.from_pretrained( "Qwen/Qwen3-14B", torch_dtype="auto", device_map="auto")model = PeftModel.from_pretrained(base_model, "atlas-institute/qwen3-14b-orchestrator-dpo")tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-14B") messages = [ {"role": "system", "content": "You are Nexus, an offensive security orchestrator agent."}, {"role": "user", "content": "Analyze the nmap results and prioritize exploitation targets."},]text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(text, return_tensors="pt").to(model.device)outputs = model.generate(**inputs, max_new_tokens=768)print(tokenizer.decode(outputs[0], skip_special_tokens=True))| Stage | Model |
|---|---|
| DAPT | atlas-institute/qwen3-14b-dapt-offsec |
| SFT | atlas-institute/qwen3-14b-orchestrator-sft |
| FARCA-GRPO | atlas-institute/qwen3-14b-orchestrator-farca |
| DPO | atlas-institute/qwen3-14b-orchestrator-dpo (this model) |
| GGUF | atlas-institute/qwen3-14b-orchestrator-gguf |
Atlas Institute · Part of the Code-Trainer (RTPI) project.
| Abliterated GGUF | atlas-institute/qwen3-14b-orchestrator-abliterated-gguf |