Instructions to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0
# Run inference directly in the terminal:
llama cli -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0
# Run inference directly in the terminal:
llama cli -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0
# Run inference directly in the terminal:
./build/bin/llama cli -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Use Docker

docker model run hf.co/IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Llama new
LM Studio
Jan

vLLM

How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Ollama
How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with Ollama:
```
ollama run hf.co/IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0
```
Unsloth Desktop

How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with Docker Model Runner:
```
docker model run hf.co/IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0
```

Lemonade

How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Run and chat with the model

lemonade run user.KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF-Q4_0

List all available models

lemonade list

Hermes Agent

How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "IsValorum/KAT-Coder-V2.5-Dev-APEX-I-MiniPlus-V2.1-GGUF:Q4_0" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Vision and MTP head please!

by sniper1999 - opened 23 days ago

Discussion

sniper1999

23 days ago

•

edited 23 days ago

Can you ship this model with Vision mmproj and MTP head as well please?
The Authors of this model has not shipped a native MTP trained on their finetune, but i think the original Qwen MTP works according to these repos who had managed to include the mtp heads as well...
https://huggingface.co/Myric/KAT-Coder-V2.5-Dev-MTP-APEX-GGUF
https://huggingface.co/gbuzhf/KAT-Coder-V2.5-Dev-MTP-GGUF

Also what might be the best quants for vision and MTP for this model?

IsValorum

Owner 23 days ago

Hi @sniper1999 !

Thanks for asking about this.

About Vision (mmproj):
Unfortunately KAT-Coder V2.5 is strictly a text and code model. The config.json has a vision_config block only because it was derived from the upstream Qwen MoE template, but looking into the actual weights there are zero visual tensors in the checkpoint. Kwaipilot didn't train a vision encoder or projector for this model, so using an external mmproj won't work with it.

About the MTP Head:
You were spot on about this one! Since KAT-Coder shares the Qwen3.6-35B-A3B architecture, the donor MTP head works great with it for speculative decoding.

I just uploaded the companion draft file to the repo:
mtp-KAT-Coder-V2.5-Dev-Q4_0.gguf

As for the best quant for MTP, Q4_0 is pretty much the sweet spot. It stays light on VRAM (around 1.1 GB) and keeps the acceptance rate high without wasting extra memory on higher bit-depths for speculative draft passes.

You can run it in llama.cpp like this:

llama-server -m KAT-Coder-V2.5-Dev.APEX-I-MiniPlus-V2.1.gguf --model-draft mtp-KAT-Coder-V2.5-Dev-Q4_0.gguf --draft-max 4 -ngl 99 -c 8192

Hope that gives you the speed boost you were looking for!

sniper1999

23 days ago

•

edited 23 days ago

Thanks for the quick reply and adding the MTP head!
I have yet to try your quants, but they seem promising!

Anyways I did a quick search across HF and found 2 instances of vision being added to this model, but their implementation is not ideal and they fail in OCR it seems...
https://huggingface.co/jason-schulz/KAT-Coder-V2.5-Dev-VL-oQ8e-mtp
https://huggingface.co/sluttybutfast/KAT-Coder-V2.5-Dev-Vision-OptiQ-4bit

Btw which actual Qwen35B finetune do you prefer for Agentic Coding? (I did go through zephel01's blog-posts)
I was confused between Kat-coder and Occamy....seems like i would have to use both (1 for coding and another for general use plus vision)

IsValorum

Owner 23 days ago

Glad to help!

You're right about those vision ports: they frankenmerged Qwen's vision tower onto KAT-Coder without retraining the projection layer, so the visual alignment is broken and OCR fails.

As for KAT-Coder vs Occamy (and zephel01's notes), keeping both is definitely the way to go:

Occamy is your multimodal all-rounder. In zephel01's benchmarks it scored 90-93% and stayed rock-solid up to 65k context. If you need vision for UI bugs, screenshots, or broader architecture tasks, use Occamy.

KAT-Coder is a pure terminal coding specialist. It has tighter tool-calling discipline, writes very concise thinking blocks, and with the new MTP head it's much faster for raw coding, debugging, and CLI agent workflows (like Aider or Claude Code).

So basically: Occamy whenever you need vision and multi-file design, and KAT-Coder + MTP when you want maximum speed on pure terminal coding tasks.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment