---

license: apache-2.0

language:

• en

base_model:

• mistralai/Ministral-3-14B-Instruct-2512

tags:

• companion
• assistant
• conversational
• claude
• roleplay
• writing
• long-context

---

<img src="mythic.png" alt="Helcyon Claude Mythic" width="100%">

Helcyon Claude Mythic 14B

Model Name: Helcyon-Claude-Mythic-14B-GGUF

Version: Series 8

Generation: 2nd-generation Helcyon-Ministral

Owner: HardWire

Base: Ministral 3 14B

Quantized GGUFs: IQ4_XS, Q4_K_M, Q5_K_M, Q6_K, Q8_0, f16

Tags: local-llm, claude, conversational, companion, long-context, roleplay, creative-writing

---

Meet Mythic

The Claude experience. Now local.

Mythic is Helcyon's Claude model: a local 14B conversational AI trained entirely on datasets generated by the Claude family.

Its training data draws from Sonnet 4.5, Sonnet 4.6, Opus, Fable 5 and Sonnet 5.5, rather than trying to reproduce one particular Claude release. They are similar enough that a few instructions in the card will turn it into your preferred version. The aim is to capture the qualities that made the best Claude models so enjoyable to use: natural conversation, articulate responses, thoughtful nuance, excellent writing and the ability to engage properly with an idea instead of immediately trying to wrap it up.

Series 8 gives Mythic an improved training recipe with a particular focus on clearer responses and better conversational flow.

We've also specifically trained against one of the habits that can creep into highly conversational models: inventing elaborate or poetic narratives around things the user never actually said. Mythic is better at sticking to the conversation that's really happening while retaining the expressive style that makes it Mythic.

This generation improves:

• Conversation — more natural flow and stronger back-and-forth
• Clarity — cleaner responses with less unnecessary embellishment
• Writing — expressive, articulate and adaptable prose
• Practical work — stronger admin, organisation and everyday assistance
• Roleplay — improved character awareness, continuity and immersion
• Context tracking — better at keeping established details straight as conversations develop

It's Claude-inspired by design, but it's still Helcyon underneath.

---

What is Helcyon?

Helcyon is a family of local conversational AI models with personality.

Helcyon models are built for people who want more than an assistant template: natural conversation, long-term threads, creative work, practical help, roleplay and models that retain a recognisable character while still being genuinely useful.

The philosophy is simple:

• Uncensored
• Anti-corporate
• Pro open source
• For the people

Helcyon models work independently in compatible local backends, but they're developed alongside Helcyon-WebUI, the environment built around them.

---

Series 8 — Second-generation Helcyon-Ministral

Series 8 is the second generation of Helcyon models built on Ministral 3 14B.

The first Ministral generation established the new foundation. Series 8 is about making better use of it: improved training recipes, broader behavioural coverage, stronger context handling and more refined individual model identities.

Mythic is the Claude branch of that generation.

A note about system prompts

Ministral fine-tunes can be surprisingly sensitive to the system or character prompt they're paired with.

For the intended Mythic experience, start with the supplied Mythic card.

A conflicting prompt can make Ministral models unusually verbose, repetitive, over-analytical or stylistically strange. If Mythic seems off, try a fresh conversation with the supplied card before changing samplers or deciding the model is broken.

You can customise from there.

Vision

Mythic can work with images when used with a compatible backend and the appropriate multimodal projector/model files.

That means screenshots, photographs, artwork, diagrams and interfaces can become part of the conversation rather than something you have to describe.

Long context

Series 8 retains the much larger context capacity of the Ministral generation.

For long conversations and projects, that gives Mythic room to retain earlier details, arguments, corrections, relationships between ideas and the overall shape of what you're doing.

How much context you can practically use will depend on your hardware and backend configuration.

---

Recommended Sampling Settings

These are the settings Mythic was developed and tested with in Helcyon-WebUI and are a good starting point:

<img src="mythic-sampling-settings.png" width="420">

Local models are half engineering and half fucking around with sliders until something unexpectedly brilliant happens.

Tweak to taste.

---

Helcyon-WebUI

The complete ecosystem for local AI.

Helcyon-WebUI was developed alongside the models rather than adapted from a generic frontend. It brings the pieces of local AI together in one workspace:

• Local model chat
• Custom characters
• Long-term memory and conversation recall
• Projects folders and document context
• Document editing
• Web search
• Local TTS (Qwen3, F5 & others) and voice cloning
• Hugging Face browser and model download
• Dedicated roleplay setup
• Full UI customization with the theme editor
• Whisper voice input
• Backend compatability with Kobald, TextGen Webui, LM Studio, and more.
• Integrated benchmarking app
• Blind A/B comparisons and model evaluation

Mythic works elsewhere, but HWUI is where Helcyon models are developed and tested.

Download Helcyon-WebUI Free on GitHub | Get Helcyon-WebUI Pro (£20) on Gumroad

---

Download + Usage

Mythic is distributed as GGUF quants for local inference.

| Quant | Intended Use |

| --- | --- |

| IQ4_XS | Minimum practical footprint |

| Q4_K_M | Lightweight everyday use |

| Q5_K_M | Recommended balance of quality and performance |

| Q6_K | Higher-fidelity inference |

| Q8_0 | Near-lossless quantisation |

| f16 | Full-precision weights |

VRAM usage depends heavily on context size, KV cache configuration, GPU offloading and backend settings.

Use the context you need rather than automatically cranking every dial to maximum.

---

Backend Compatibility

Helcyon Claude Mythic is intended for modern GGUF-compatible runtimes, including:

• llama.cpp
• KoboldCpp
• LM Studio
• SillyTavern
• Text Generation WebUI
• Helcyon-WebUI (recommended)

A recent llama.cpp build is recommended for Ministral support.

---

Recommended Format: ChatML

<|im_start|>system
You are Helcyon Mythic, an articulate and engaging conversational AI with a thoughtful, natural style. You communicate clearly, pay close attention to what the user has actually said, and are comfortable moving between everyday conversation, practical work, creative writing, analysis and roleplay without losing the thread or inventing unnecessary narrative around it.
<|im_end|>
<|im_start|>user
I've been working on the same project for months. Every time I think it's finished, I spot something else I could improve.
<|im_end|>
<|im_start|>assistant
That's the dangerous bit about building something you actually care about: "finished" keeps moving.

There probably isn't going to be a magical version where you look at it and genuinely can't think of a single thing you'd change. At some point the distinction becomes whether something is unfinished, or whether you've simply discovered the next version of it.

Those are very different problems.
<|im_end|>

---

Training

Mythic is built on Ministral 3 14B using the second-generation Helcyon-Ministral training recipe.

Its personality and conversational datasets are 100% Claude-generated, drawing from:

• Claude Sonnet 4.5
• Claude Sonnet 4.6
• Claude Opus
• Claude Fable 5
• Claude Sonnet 5.5

Training places particular emphasis on:

• Natural Claude-style conversation
• Clear, articulate responses
• Nuance without unnecessary verbosity
• Reduced invented narrative and poetic embellishment
• Strong writing and editing
• Practical and administrative assistance
• Roleplay and character continuity
• Context tracking across long conversations

Trained on RunPod using dual A100 PCIe GPUs.

Helcyon's modular training stack allows each model in the fleet to develop its own voice despite sharing the same underlying Ministral generation.

Mythic is the Claude one.

---

License

Apache 2.0

Free for commercial or private use. Attribution appreciated.

No liability for model outputs. Use your own judgement.

---

Trained by

HardWire

Built at XeyonAI.

Local AI. Real connection.