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.

I'd love your feedback—have you tried this? How does it perform for you?

Click here to get info on choosing the right GGUF model format

katanemo/Arch-Agent-3B

Overview

Arch-Agent is a collection of state-of-the-art (SOTA) LLMs specifically designed for advanced function calling and agent-based applications. Designed to power sophisticated multi-step and multi-turn workflows, Arch-Agent excels at handling complex, multi-step tasks that require intelligent tool selection, adaptive planning, and seamless integration with external APIs and services. Built with a focus on real-world agent deployments, Arch-Agent delivers leading performance in complex scenarios while maintaining reliability and precision across extended function call sequences. Key capabilities inlcude:

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.

For more details, including fine-tuning, inference, and deployment, please refer to our Github.

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.

[!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.

Requirements

The code of Arch-Agent-3B has been in the Hugging Face transformers library and we recommend to install latest version:

pip install transformers>=4.51.0

How to use

We use the following example to illustrate how to use our model to perform function calling tasks. Please note that, our model works best with our provided prompt format. It allows us to extract JSON output that is similar to the OpenAI's function calling.

Quickstart

python

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

model_name = "katanemo/Arch-Agent-3B"

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)

License

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

🚀 If you find these models useful

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 unfortunatly 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 lastest models hosted on Novita.

💡 Example commands you could test:

"Give me info on my websites SSL certificate"
"Check if my server is using quantum safe encyption for communication"
"Run a comprehensive security audit on my server"
'"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!

Final Word

I fund the servers used to create these model files, run the Quantum Network Monitor service, and pay for inference from Novita and OpenAI—all out of my own pocket. All the code behind the model creation and the Quantum Network Monitor project is open source. Feel free to use whatever you find helpful.

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! 😊

Mungert/Arch-Agent-3B-GGUF

Arch-Agent-3B GGUF Models

Model Generation Details

Quantization Beyond the IMatrix