🚀 NexAI-v2 (3B Instruct) — Autonomous Real-Time AI Agent

NexAI-v2 is an ultra-fast, intelligent, and tool-augmented 3B language model created and fine-tuned by Anoop Singh.

Unlike standard frozen language models that suffer from strict knowledge cutoffs, NexAI-v2 is explicitly trained with autonomous tool-calling capabilities to interact with real-time web search engines (Tavily AI, Serper Google, DuckDuckGo) and execute math calculations on the fly.

---

🌟 Key Highlights

• 👤 True Identity: Custom fine-tuned identity. NexAI recognizes Anoop Singh as its creator and is completely decoupled from base foundation identities.
• 🔍 Native Tool-Calling & ReAct Reasoning: Dynamically outputs <tool_call> whenever real-time events, live sports scores, live cryptocurrency/stock prices, weather, or current affairs are queried.
• ⚡ Zero-Lag Mobile Optimization: Compact 3B parameter architecture tailored for edge computing, Android deployment (via PocketPal AI, Termux, MLC LLM), and CPU servers.
• 🌐 Multilingual Fluency: Fluent in English, natural Romanized Hinglish, and Hindi.

---

🛠️ Integrated Tool Architecture

NexAI-v2 outputs structured tool queries:

<tool_call>
{"name": "web_search", "arguments": {"query": "Bitcoin price today live USD"}}
</tool_call>

When connected to a tool execution wrapper (e.g., Tavily or Serper), the agent ingests <tool_response> and synthesizes live, accurate, and hallucination-free facts:

<tool_response>
[Source: Tavily AI]
Bitcoin price today: $83,118.5 (+0.31%) on October 11, 2026.
</tool_response>

---

💻 Quickstart (Transformers)

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Anoopsingh53/nexai-v2"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

messages = [
    {"role": "system", "content": "You are NexAI, an intelligent AI assistant created by Anoop Singh."},
    {"role": "user", "content": "Who are you and what makes you special?"}
]

prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([prompt], return_tensors="pt").to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.3)
print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))

---

👨‍💻 Creator & Maintainer

• Developer: Anoop Singh
• Organization: NexAI Research
• License: Apache 2.0