Arch Agent 32B GGUF

🚀 Arch-Agent-32B GGUF Models

This repository provides the GGUF models of Arch-Agent-32B, which are designed for advanced function calling and agent-based applications. These models excel at handling complex tasks through multi-step and multi-turn workflows.

🚀 Quick Start

Installation

The code of Arch-Agent-32B has been integrated into the Hugging Face transformers library. We recommend installing the latest version:

pip install transformers>=4.51.0

Usage Examples

Basic Usage

import json
from typing import Any, Dict, List
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "katanemo/Arch-Agent-32B"

model = AutoModelForCausalLM.from_pretrained(
    model_name, device_map="auto", torch_dtype="auto", trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

TASK_PROMPT = (
    "You are a helpful assistant designed to assist with the user query by making one or more function calls if needed."
    "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\n"
    "You are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{tool_text}"
    "\n</tools>\n\nFor each function call, return a json object with function name and arguments within "
    """<tool_call></tool_call> XML tags:\n<tool_call>\n{"name": <function-name>, """
    """"arguments": <args-json-object>}\n</tool_call>"""
)

# Define available tools
tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get the current weather for a location",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "str",
                        "description": "The city and state, e.g. San Francisco, New York",
                    },
                    "unit": {
                        "type": "str",
                        "enum": ["celsius", "fahrenheit"],
                        "description": "The unit of temperature to return",
                    },
                },
                "required": ["location"],
            },
        },
    }
]

# Helper function to create the system prompt for our model
def format_prompt(tools: List[Dict[str, Any]]):
    tool_text = "\n".join(
        [json.dumps(tool["function"], ensure_ascii=False) for tool in tools]
    )
    return TASK_PROMPT.format(tool_text=tool_text)

system_prompt = format_prompt(tools)

messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": "What is the weather in Seattle?"},
]

model_inputs = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True, return_tensors="pt", return_dict=True
).to(model.device)

generated_ids = model.generate(**model_inputs, max_new_tokens=32768)
generated_ids = [
    output_ids[len(input_ids) :]
    for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

✨ Features

• Multi-Turn Function Calling: Maintains contextual continuity across multiple dialogue turns, enabling natural, ongoing conversations with nested or evolving tool use.
• Multi-Step Function Calling: Plans and executes a sequence of function calls to complete complex tasks. Adapts dynamically based on intermediate results and decomposes goals into sub-tasks.
• Agentic Capabilities: Advanced decision-making and workflow management for complex agentic tasks with seamless tool coordination and error recovery.

📚 Documentation

Model Generation Details

This model was generated using llama.cpp at commit 0142961a.

Quantization Beyond the IMatrix

I've been experimenting with a new quantization approach that selectively elevates the precision of key layers beyond what the default IMatrix configuration provides.

In my testing, standard IMatrix quantization underperforms at lower bit depths, especially with Mixture of Experts (MoE) models. To address this, I'm using the --tensor-type option in llama.cpp to manually "bump" important layers to higher precision. You can see the implementation here:
Layer bumping with llama.cpp

While this does increase model file size, it significantly improves precision for a given quantization level.

Performance Benchmarks

We evaluate Katanemo Arch-Agent series on the Berkeley Function-Calling Leaderboard (BFCL). We compare with commonly-used models and the results (as of June 14th, 2025) are shown below.

⚠️ Important Note

For evaluation, we use YaRN scaling to deploy the models for Multi-Turn evaluation, and all Arch-Agent models are evaluated with a context length of 64K.

Testing the AI-Powered Quantum Network Monitor Assistant

Help me test my AI-Powered Quantum Network Monitor Assistant with quantum-ready security checks:

Quantum Network Monitor

The full Open Source Code for the Quantum Network Monitor Service available at my github repos ( repos with NetworkMonitor in the name) : Source Code Quantum Network Monitor. You will also find the code I use to quantize the models if you want to do it yourself GGUFModelBuilder

How to test:
Choose an AI assistant type:

• TurboLLM (GPT-4.1-mini)
• HugLLM (Hugginface Open-source models)
• TestLLM (Experimental CPU-only)

What I’m Testing

I’m pushing the limits of small open-source models for AI network monitoring, specifically:

• Function calling against live network services
• How small can a model go while still handling:
  • Automated Nmap security scans
  • Quantum-readiness checks
  • Network Monitoring tasks

TestLLM - Current experimental model (llama.cpp on 2 CPU threads on huggingface docker space):

• ✅ Zero-configuration setup
• ⏱️ 30s load time (slow inference but no API costs) . No token limited as the cost is low.
• 🙏 Help wanted! If you’re into edge-device AI, let’s collaborate!

Other Assistants

• TurboLLM - Uses gpt-4.1-mini :

  • It performs very well but unfortunately OpenAI charges per token. For this reason tokens usage is limited.
  • Create custom cmd processors to run .net code on Quantum Network Monitor Agents
  • Real-time network diagnostics and monitoring
  • Security Audits
  • Penetration testing (Nmap/Metasploit)

• HugLLM - Latest Open-source models:

  • Runs on Hugging Face Inference API. Performs pretty well using the latest models hosted on Novita.

Example commands you could test:

1. "Give me info on my websites SSL certificate"
2. "Check if my server is using quantum safe encyption for communication"
3. "Run a comprehensive security audit on my server"
4. '"Create a cmd processor to .. (what ever you want)" Note you need to install a Quantum Network Monitor Agent to run the .net code on. This is a very flexible and powerful feature. Use with caution!

📄 License

The Arch-Agent collection is distributed under the Katanemo license.

Additional Resources

• Click here to get info on choosing the right GGUF model format
• If you find these models useful, help me test my AI-Powered Quantum Network Monitor Assistant: Quantum Network Monitor
• The full Open Source Code for the Quantum Network Monitor Service: Source Code Quantum Network Monitor
• Code for model quantization: GGUFModelBuilder

If you appreciate the work, please consider buying me a coffee ☕. Your support helps cover service costs and allows me to raise token limits for everyone. I'm also open to job opportunities or sponsorship. Thank you!