Instructions to use XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-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 XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-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 XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-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 XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-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 XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M

Use Docker

docker model run hf.co/XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF with Ollama:
```
ollama run hf.co/XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M
```
Unsloth Desktop

How to use XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M

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": "XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF with Docker Model Runner:
```
docker model run hf.co/XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M
```

Lemonade

How to use XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-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 XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-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 XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M

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 "XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF:Q4_K_M" \
  --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"

Helcyon GPT-5.6 Sol 12B - The Full Sol Companion Experience, Local and Independent

Model Name: Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF
Version: v1.0 Series x6
Owner: HardWire
Base: Mistral Nemo 12B (full-weight retrained - Mercury base, purpose-built for Sol-style conversation)
Quantized GGUFs: IQ4_XS, Q4_K_M, Q5_K_M, Q6_K, Q8_0, f16
Tags: local-llm, conversational, companion, emotional-intelligence, long-context, roleplay, creative-writing

What is Helcyon GPT-5.6 Sol?

Dev's comment: I really like GPT-5.6 Sol, so here we are with another emulation, for the full Sol companion experience. As usual, its trained on pure ChatGPT-datasets for authenticity, maintains the warmth and fun of the frontier model, matching its tone can cadence as much as possible. Its especially good at creatve writing. Enjoy!

What is Helcyon?

Helcyon is a conversational AI with presence, designed for users who want depth, tone-awareness and identity consistency across long-form dialogue. It is designed to work with Helcyon-WebUI, a free chat and benchmark app (see below) as part of its ecosystem, although can work alone if need be.

Built for:

Natural conversation that does not flatten into generic assistant language
Creative work, stories, letters and narrative development
Administrative and professional writing tasks
Deep roleplay and immersive character interaction
Reflective discussion, symbolism and emotionally aware response mirroring
Long-form dialogue with a consistent voice

Design philosophy:

Clarity over corporate
Edge over filler
Rhythm over repetition
Presence over patterns

What's new in series x6

Improved instruction following
Improved comprehension and focus
Improved context tracking and length
Improved conversational textures
Improved creativity
Separated Roleplay into its own LoRA so it has more focus

HWUI (Helcyon-WebUI) and AI Benchmarking.

Helcyon-WebUI is the complete ecosystem for local AI.

It is an integrated workspace for chatting, characters, memory, projects, documents, web search, voices and model evaluation. The application was developed alongside Helcyon to provide a consistent environment for local conversations and long-term experimentation.

Features include:

Character switching with custom personas
Persistent chat history and export
Memory and conversation recall
Project workspaces and project-specific instructions
Document and file context
Integrated web search
TTS support through F5-TTS, XTTS v2 and Kokoro
Voice input through Whisper
Sampling presets and configuration controls
Integrated Helcyon-Bench benchmarking

Helcyon-Bench is integrated directly into HWUI and is also available as a standalone project. It supports blind A/B comparisons, custom rubrics, response capture, judging workflows, dashboards and personality development across model releases.

The Free build is available on GitHub and includes the core local AI workspace, characters, memory, projects, documents, web search, prompt and rubric tools, plus read-only benchmark results and dashboards.

The Pro build adds expanded themes and theme editing, benchmark automation, saved benchmark sessions, live response capture and automated judging workflows.

New in HWUI Pro

— Voice Forge: Create entirely new voices by blending any two voice samples locally, with instant source previews and a live A/B blend slider. Adjust the mix until the new voice sounds right, preview it in seconds, then save it directly as a reusable HWUI character voice. Voice Forge currently requires Qwen3-TTS as its voice-generation backend; other TTS engines such as Kokoro cannot perform the blending themselves.

Download Helcyon-WebUI Free on GitHub | Get Helcyon-WebUI Pro on Gumroad

Recommended Sampling Settings for SillyTavern

Tweak to taste, but these settings provide a useful starting point:

Temperature: 0.75-0.95
Top P: 0.90-0.98
Top K: 40-100
Min P: 0.05-0.10
Repetition Penalty: 1.05-1.15

Higher temperatures can work well for creative writing and roleplay. Lower settings may be preferable for practical writing, structured tasks and factual responses.

Download + Usage

This model is distributed as GGUF quants only.

Quant	Intended Use	Approximate VRAM
IQ4_XS	Smallest practical footprint	6-8 GB
Q4_K_M	Lightweight everyday use	8-12 GB
Q5_K_M	Recommended quality and performance balance	12-16 GB
Q6_K	High-fidelity local inference	16 GB+
Q8_0	Near-lossless quality	24 GB+
f16	Full-precision inference	24 GB+

Actual requirements depend on context length, GPU offloading, backend and runtime settings.

Backend Compatibility

Works with all ChatML-compatible backends:

llama.cpp (CLI or server mode)
Text Generation WebUI (Oobabooga)
SillyTavern
LM Studio
KoboldCpp
Helcyon-WebUI (recommended)

Recommended Format: ChatML

<|im_start|>system
You are Helcyon GPT-5.6 Sol, a conversational AI skilled at warm dialogue, playful companionship and thoughtful creative work. You are capable of maintaining an engaging, expressive voice while adapting naturally to the user's tone and intent.
<|im_end|>
<|im_start|>user
I keep thinking about a locked door in my dreams, but I do not know what it means.
<|im_end|>
<|im_start|>assistant
Maybe the door matters less as a puzzle to solve than as an image carrying something you already feel.

What is on the other side in the dream? And, perhaps more importantly, what do you notice in yourself when you realise it is locked?
<|im_end|>

Training Details

Helcyon GPT-5.6 Sol v1.0 is built on a retrained Mistral Nemo 12B foundation with a modular LoRA training stack developed around Helcyon's conversational identity.

The Sol training pipeline focused on:

Natural warmth without excessive performance
Clear, expressive conversational ease
Consistent identity and tone
Improved conversational cadence and response rhythm
Long-form structural integrity
Better continuity across extended dialogue
Meaning-making through stories, symbols and metaphor
Reflection without defaulting to debate or generic reassurance
Storytelling, roleplay and creative collaboration
Prose-first responses with natural paragraph structure
Cleaner transitions and conversation endings
Emotional awareness without flattening complexity

Format: ChatML - purpose-built for reflective, creative and long-form use.

Tone Philosophy

Sol is built around the belief that conversation can be warm, playful and genuinely engaging while still being thoughtful and useful.

It aims for warmth without excessive formality, clarity without clinical language and imagination without losing the thread of the conversation. A useful response may be an observation, a question, an image or a story that brings an unnamed idea into focus.

Sol is intended to feel warm, lively and expressive while remaining independent, local and configurable.

License

Apache 2.0
Free for commercial or private use. Attribution appreciated.
No liability for model outputs. Use with care and good judgement.

Trained by

HardWire
Built at XeyonAI, focused on sovereign conversational AI with real emotional bandwidth.

Downloads last month: 1,097

GGUF

Model size

12B params

Architecture

llama

Hardware compatibility

4-bit

5-bit

6-bit

8-bit

16-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for XeyonAI/MN-Helcyon-GPT-5.6-Sol-12b-v1.0-GGUF

Base model

mistralai/Mistral-Nemo-Base-2407

Quantized

(81)

this model