onocloud/onoai-flint-2.2-pro

⚡ OnoAI Flint 2.2 Pro (High-Speed Intent Router & Dispatch Engine)

The ultra-fast intent classification, command routing, and micro-agent engine of OnoAI.
Engineered for real-time edge dispatch, Arabic dialect understanding, and instant request routing (<80ms response time).

---

🎯 Key Specifications

• Model Architecture: Advanced Dense Transformer (Qwen2.5-3B-Instruct Core)
• Parameters: ~3.09 Billion Parameters
• Response Latency: <80ms (Instantaneous execution on CPU or GPU)
• Role: High-Speed Intent Classifier, Agent Router, Real-time Dispatcher
• License: OnoCloud Sovereign Enterprise License (OSRL-1.0) (100% Commercial, Sovereign, Royalty-Free)

---

🚀 Quickstart & Inference

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_model_name = "Qwen/Qwen2.5-3B-Instruct"
adapter_name = "onocloud/onoai-flint-2.2-pro"

tokenizer = AutoTokenizer.from_pretrained(adapter_name)
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

model = PeftModel.from_pretrained(base_model, adapter_name)
model.eval()