Instructions to use gdfhhjk/spectrida-re-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use gdfhhjk/spectrida-re-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="gdfhhjk/spectrida-re-gguf",
	filename="spectrida-re-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use gdfhhjk/spectrida-re-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf gdfhhjk/spectrida-re-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf gdfhhjk/spectrida-re-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf gdfhhjk/spectrida-re-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf gdfhhjk/spectrida-re-gguf:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf gdfhhjk/spectrida-re-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf gdfhhjk/spectrida-re-gguf:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf gdfhhjk/spectrida-re-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf gdfhhjk/spectrida-re-gguf:Q4_K_M

Use Docker

docker model run hf.co/gdfhhjk/spectrida-re-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use gdfhhjk/spectrida-re-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gdfhhjk/spectrida-re-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": "gdfhhjk/spectrida-re-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/gdfhhjk/spectrida-re-gguf:Q4_K_M

Ollama
How to use gdfhhjk/spectrida-re-gguf with Ollama:
```
ollama run hf.co/gdfhhjk/spectrida-re-gguf:Q4_K_M
```

Unsloth Studio

How to use gdfhhjk/spectrida-re-gguf with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for gdfhhjk/spectrida-re-gguf to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for gdfhhjk/spectrida-re-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for gdfhhjk/spectrida-re-gguf to start chatting

How to use gdfhhjk/spectrida-re-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf gdfhhjk/spectrida-re-gguf:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "gdfhhjk/spectrida-re-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use gdfhhjk/spectrida-re-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf gdfhhjk/spectrida-re-gguf:Q4_K_M

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 gdfhhjk/spectrida-re-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use gdfhhjk/spectrida-re-gguf with Docker Model Runner:
```
docker model run hf.co/gdfhhjk/spectrida-re-gguf:Q4_K_M
```

Lemonade

How to use gdfhhjk/spectrida-re-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull gdfhhjk/spectrida-re-gguf:Q4_K_M

Run and chat with the model

lemonade run user.spectrida-re-gguf-Q4_K_M

List all available models

lemonade list

Spectrida-RE 🧠🔬

"Yo dawg, I heard you like reverse engineering, so I put an RE model in your RE workflow so you can reverse while you reverse."

An 8B parameter Qwen3 model fine-tuned specifically for reverse engineering tasks — reading assembly, naming functions, detecting virtual machines, and understanding obfuscated code. Basically it's like having a junior RE that doesn't ask for coffee breaks.

🎉 SHOUTOUT TO THE 124+ DOWNLOADERS

Y'all are legends. 124 downloads in a day with zero marketing? That's more attention than my high school crush ever gave me.

Seriously though — if you're reading this and you downloaded the model, thank you! 🙏 You could've been scrolling TikTok but instead you chose to download an 8.7GB file from some random person on the internet. That takes trust. Or terrible impulse control. Either way, I appreciate it.

If the model works for you, cool. If it doesn't, no hard feelings — just blame the GGUF quantization, not me.

Training Approach

Neuron-Level Targeting

The model is probed with assembly/RE questions to identify which neurons activate
Only activated neurons are fine-tuned using Unsloth
Prevents catastrophic forgetting and makes training highly efficient
Think of it as surgery vs. hitting your brain with a hammer

Two-Stage Training

SFT (Supervised Fine-Tuning): Applied only to targeted neurons
GRPO (Group Relative Policy Optimization): Uses binary rewards (correct/incorrect) + confidence scoring

We tried stage 3 (good vibes and positive reinforcement) but the model didn't respond. Rude.

Hardware

Trained on an RTX 4070 + Ryzen 5800X3D with 32GB RAM. No datacenters were harmed in the making of this model.

Tool Integration

Works with IDA Pro via MCP tools (100% reliability — yes, we tested)
Connects to a local context server for extended context
Trained on TigressVM obfuscation patterns because we hate ourselves

Capabilities

Function Naming

Successfully named ~50,000 functions in Mario Odyssey's main.nso
Context-aware naming based on cross-references, strings, calling conventions
Still can't figure out what sub_123456 does though. Some mysteries are eternal.

VM Detection

Identifies virtual machine obfuscation patterns
Trained on TigressVM specifically
Detection rate: better than your average antivirus, worse than your paranoid friend

Obfuscation Handling

Follows function combinations and traces execution
Understands big-picture context from disassembly traces
Gets confused by the same things you do. Welcome to RE, baby.

Deployment

GGUF format — runs on CPU, phone, Raspberry Pi
Designed to work alongside a modded IDA Pro using Capstone + multi-threading (4 hours → 67 seconds for Among Us binary)
Yes, it can run on your phone. No, I don't know why you'd want to disassemble things on your phone. But you do you.

🔮 Future: Spectrida Suite — The RE Terminal That Didn't Need to Exist

This model is just the brain. The body is coming soon.

I'm building a full TUI (Terminal User Interface) that brings together everything into one beautiful chaos:

🧠 MCP Servers (Model Context Protocol)

This model was trained specifically with these MCP servers. It expects them. Feed it something else and it gets confused. Don't be that person.

Spectrida-RE MCP — query the model directly
mrexodia/ida-pro-mcp — the OG (9k+ ⭐)
jtsylve/ida-mcp — headless IDA with idalib
unrealsoftwaredev/ida-mcp — 80+ tools across 12 domains
blacktop/ida-mcp-rs — Rust IDA MCP
DeusData/codebase-memory-mcp — context server (knowledge graph, code intelligence)
Ollama MCP — local model swapping

⚠️ Trained on these. If your MCP setup doesn't match, expect bad function names and existential dread.

🛠️ Built-in Tools

Binary toolchain — hex viewer, string extractor, file type detection, all the good RE stuff
🐅 TigressVM Analyzer — dedicated module for Tigress obfuscation detection and deobfuscation
⛓️ Call Chain Tracing — follow function call chains, map execution flow, understand the big picture
📝 Notes — save findings, tag them, don't lose them in a random text file again

Basically a RE-focused terminal app that doesn't make you juggle 15 windows like a circus performer. One screen. All the power. Maximum chaos.

The Tigress and call chain modules are still cooking — RE is hard, ok? But when they're done, this thing's gonna eat obfuscated binaries for breakfast.

GitHub repo coming when it's ready. Follow the model for updates or just manifest it into existence by downloading the GGUF again. Every download adds 0.01% faster development speed. (Not true, but let's see how many of you read this far.)

Why "Spectrida"?

Because "ReverseEngineeringModelThatNamesFunctionsAndDetectsVMs" didn't fit on the disk.

License

Apache 2.0 — do whatever you want, just don't blame me when it names a function definitely_not_a_backdoor() and it turns out to be exactly that.

Built with ❤️, ☕, and a concerning amount of spite.

If you read this far, you're legally obligated to star the future repo. I don't make the rules.

Downloads last month: 268

GGUF

Model size

8B params

Architecture

qwen3

Hardware compatibility

4-bit

Model tree for gdfhhjk/spectrida-re-gguf

Base model

Qwen/Qwen3-8B-Base

Finetuned

Qwen/Qwen3-8B

Quantized

(306)

this model