Downloads · 30 days
2.3K
30% of all-time downloads
Gemma-4-Gembrain-X-Core-31B-GGUF
Nimbz/Gemma-4-Gembrain-X-Core-31B-GGUF
Gemma-4-Gembrain-X-Core-31B-GGUF is a machine learning model from Nimbz. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
<style .gem-wrapper { background-color: 06090a; background-image: radial-gradient(circle at 16% 12%, rgba(16, 185, 129, 0.15) 0%, transparent 42%), radial-gradient(circle at 84% 86%, rgba(45, 212, 191, 0.13) 0%, trans…
All-time downloads
7.8K
Public
Repo size
227 GB
Likes
6
Public
Repo makeup
Click a slice to open those files.
.gguf171 GB · 100%
At a glance
- License
- apache-2.0
- Access
- Public
- Created
- Jun 30, 2026
- Updated
- Jun 30, 2026
- SHA
- 53c11486
Base models
Model card
From the Hugging Face model README
<style>
.gem-wrapper {
background-color: #06090a;
background-image:
radial-gradient(circle at 16% 12%, rgba(16, 185, 129, 0.15) 0%, transparent 42%),
radial-gradient(circle at 84% 86%, rgba(45, 212, 191, 0.13) 0%, transparent 46%),
radial-gradient(circle at 50% 28%, rgba(52, 211, 153, 0.11) 0%, transparent 56%),
radial-gradient(circle at 78% 22%, rgba(163, 230, 53, 0.07) 0%, transparent 48%),
radial-gradient(circle at 24% 76%, rgba(251, 191, 36, 0.07) 0%, transparent 46%),
radial-gradient(1.4px 1.4px at 22% 30%, rgba(110,231,183,0.7), transparent),
radial-gradient(1.3px 1.3px at 66% 18%, rgba(251,191,36,0.6), transparent),
radial-gradient(1.5px 1.5px at 82% 40%, rgba(94,234,212,0.6), transparent),
radial-gradient(1.2px 1.2px at 36% 60%, rgba(251,191,36,0.5), transparent),
radial-gradient(1.3px 1.3px at 72% 66%, rgba(110,231,183,0.55), transparent),
radial-gradient(1.2px 1.2px at 48% 86%, rgba(94,234,212,0.5), transparent),
radial-gradient(1.1px 1.1px at 12% 50%, rgba(163,230,53,0.5), transparent);
color: #cbd5d3;
font-family: 'Iowan Old Style', 'Palatino Linotype', Georgia, 'Times New Roman', serif;
padding: 42px;
border-radius: 6px;
border: 1px solid #134e4a;
line-height: 1.75;
position: relative;
overflow: hidden;
box-shadow: inset 0 0 130px rgba(0,0,0,0.92), 0 20px 55px rgba(0,0,0,0.6);
}
.gem-wrapper::before {
content: "⬡";
position: absolute;
top: -34px; right: -10px;
font-size: 200px;
opacity: 0.06;
color: #10b981;
filter: drop-shadow(0 0 40px rgba(16,185,129,0.55));
pointer-events: none;
z-index: 0;
}
.gem-header { text-align: center; padding-bottom: 26px; position: relative; z-index: 10; }
.gem-title {
font-size: 2.0rem;
font-weight: 800;
margin: 18px 0 6px 0;
letter-spacing: 10px;
background: linear-gradient(90deg, #059669, #10b981, #2dd4bf, #a3e635, #34d399, #fbbf24);
-webkit-background-clip: text;
background-clip: text;
-webkit-text-fill-color: transparent;
filter: drop-shadow(0 0 20px rgba(16, 185, 129, 0.5));
}
.gem-sub {
color: #34d399;
text-transform: uppercase;
letter-spacing: 7px;
font-size: 0.82rem;
text-shadow: 0 0 12px rgba(16, 185, 129, 0.6);
}
.gem-insane {
color: #fbbf24;
text-transform: uppercase;
letter-spacing: 6px;
font-size: 0.92rem;
margin-top: 10px;
text-shadow: 0 0 14px rgba(251, 191, 36, 0.5);
}
.gem-tagline { color: #94a3a0; font-style: italic; font-size: 1.15rem; letter-spacing: 1px; margin-top: 14px; }
.stat-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 16px; margin: 28px 0 6px; position: relative; z-index: 5; }
.stat-item {
text-align: center;
padding: 16px 10px;
background: linear-gradient(180deg, #0c1413 0%, #07090a 100%);
border: 1px solid #134e4a;
border-radius: 4px;
box-shadow: inset 0 2px 12px rgba(16, 185, 129, 0.05), 0 8px 22px rgba(0,0,0,0.45);
transition: all 0.25s ease;
}
.stat-item:hover { border-color: #10b981; box-shadow: inset 0 2px 12px rgba(16,185,129,0.12), 0 0 22px rgba(251,191,36,0.28); transform: translateY(-2px); }
.stat-value { display: block; font-size: 1.05rem; font-weight: bold; letter-spacing: 2px; color: #e6efe9; text-shadow: 0 0 12px rgba(45, 212, 191, 0.45); }
.ritual-card {
background: linear-gradient(145deg, rgba(11, 20, 18, 0.9), rgba(7, 9, 10, 0.96));
border: 1px solid #173f37;
border-left: 5px solid #10b981;
padding: 26px 30px;
margin: 24px 0;
border-radius: 4px 16px 16px 4px;
box-shadow: inset 0 -2px 12px rgba(16, 185, 129, 0.05), 0 10px 30px rgba(0,0,0,0.55);
position: relative;
z-index: 5;
}
.ritual-card h2, .ritual-card h3 { color: #34d399; margin-top: 0; letter-spacing: 2px; }
.gem-prism-card {
border-left-color: #fbbf24;
border-color: #4d3a10;
background: linear-gradient(145deg, rgba(26, 20, 8, 0.9), rgba(8, 8, 7, 0.97));
box-shadow: inset 0 -2px 14px rgba(251, 191, 36, 0.08), 0 0 26px rgba(16, 185, 129, 0.16);
}
.gem-prism-card h3 { color: #fcd34d; }
.gem-h2 {
color: #e6efe9;
letter-spacing: 4px;
border-bottom: 1px solid #173f37;
padding-bottom: 10px;
position: relative;
z-index: 5;
margin-top: 38px;
}
.gem-h2 .glyph { color: #10b981; text-shadow: 0 0 10px rgba(16,185,129,0.6); }
hr.facet {
border: 0; height: 1px;
background-image: linear-gradient(to right, rgba(0,0,0,0), #059669, #10b981, #2dd4bf, #a3e635, #fbbf24, rgba(0,0,0,0));
margin: 18px 0;
}
.gem-pillars { display: grid; grid-template-columns: repeat(auto-fit, minmax(220px, 1fr)); gap: 16px; margin: 18px 0; position: relative; z-index: 5; }
.pillar {
background: linear-gradient(160deg, rgba(12, 21, 19, 0.92), rgba(7, 9, 10, 0.97));
border: 1px solid #173f37;
border-top: 3px solid #10b981;
border-radius: 6px;
padding: 18px 20px;
box-shadow: 0 8px 24px rgba(0,0,0,0.5);
}
.pillar h4 { margin: 0 0 6px; color: #34d399; letter-spacing: 2px; font-size: 1.0rem; }
.pillar .pstar { color: #fbbf24; }
.pillar p { margin: 0; font-size: 0.9rem; color: #94a3a0; }
.gem-table { width: 100%; border-collapse: collapse; background: rgba(0,0,0,0.32); border-radius: 6px; overflow: hidden; border: 1px solid rgba(16, 185, 129, 0.2); margin-top: 16px; }
.gem-table td, .gem-table th { padding: 11px 18px; border-bottom: 1px solid rgba(16, 185, 129, 0.12); font-size: 0.92rem; }
.gem-table th { text-align: left; background: rgba(19, 78, 74, 0.55); color: #e6efe9; border-bottom: 2px solid #134e4a; letter-spacing: 2px; }
.gem-table tr:last-child td { border-bottom: none; }
.gem-table td:first-child { color: #6ee7b7; font-weight: 500; }
.gem-table tr:hover { background: rgba(16, 185, 129, 0.06); }
.gem-gtable { display: grid; grid-template-columns: max-content 1fr; width: 100%; background: rgba(0,0,0,0.32); border: 1px solid rgba(16, 185, 129, 0.2); border-radius: 6px; overflow: hidden; }
.gem-gtable > div { padding: 11px 18px; border-bottom: 1px solid rgba(16, 185, 129, 0.12); font-size: 0.92rem; }
.gem-gtable .vg-head { background: rgba(19, 78, 74, 0.55); color: #e6efe9; letter-spacing: 2px; font-weight: bold; border-bottom: 2px solid #134e4a; }
.gem-gtable .vg-cell { color: #94a3a0; font-weight: 500; }
.gem-gtable .vg-val { text-align: right; color: #fbbf24; font-family: monospace; font-size: 1.05em; }
.gem-gtable > div:nth-last-child(-n+2) { border-bottom: none; }
.gem-btns { text-align: center; margin: 6px 0 4px; position: relative; z-index: 5; }
.gem-btn {
display: inline-block;
font-size: 0.95rem; letter-spacing: 2px; font-weight: bold;
color: #e6efe9 !important;
text-decoration: none !important;
background: linear-gradient(135deg, #0c1413, #07090a);
border: 1px solid #10b981;
padding: 13px 30px; margin: 7px;
border-radius: 4px;
box-shadow: 0 0 14px rgba(16, 185, 129, 0.26);
transition: all 0.25s ease;
}
.gem-btn:hover { background: linear-gradient(135deg, #134e4a, #115e4a); box-shadow: 0 0 22px rgba(251,191,36,0.45); transform: translateY(-2px); }
.gem-btn.special { background: linear-gradient(135deg, #134e4a, #3d2f0c); border-color: #fbbf24; box-shadow: 0 0 18px rgba(251, 191, 36, 0.4); }
.note-box { background: rgba(0,0,0,0.5); padding: 15px 18px; border-radius: 6px; border-left: 3px solid #fbbf24; margin-top: 16px; color: #cbd5d3; font-size: 0.92rem; }
.note-box strong { color: #34d399; }
.gem-credits ul { list-style: none; padding-left: 0; }
.gem-credits li { margin-bottom: 13px; position: relative; padding-left: 28px; color: #cbd5d3; }
.gem-credits li::before {
content: "◈";
position: absolute; left: 0; top: 0;
font-size: 0.9em;
color: #fbbf24;
filter: drop-shadow(0 0 6px rgba(251, 191, 36, 0.7));
}
.gem-credits a, .gem-wrapper a { color: #6ee7b7; text-decoration: none; border-bottom: 1px dotted #6ee7b7; }
.gem-credits a:hover, .gem-wrapper a:hover { color: #34d399; border-color: #34d399; }
.crystal-rule { text-align: center; color: #173f37; letter-spacing: 12px; margin: 26px 0 4px; }
.summary-style summary { color: #34d399; cursor: pointer; font-weight: bold; letter-spacing: 1px; padding: 6px 0; }
</style>
<div class="gem-wrapper">
<div class="ritual-card gem-prism-card">
<div>◇ <b style="color:#fcd34d;">NOTE:</b> After initial feedback from the BeaverAI Community, I kicked the last abliteration step. This should help with preserving the "soul" of the merge, making it less blunt and less agreeable. Awaiting more feedback.</div>
</div>
<div class="gem-header">
<img src="gembrain-x-core-header.png" width="90%" style="border-radius: 6px; border: 1px solid #173f37; box-shadow: 0 15px 45px rgba(0,0,0,0.7); margin: auto; margin-bottom: 14px;">
<hr class="facet">
<h1 class="gem-title">
<div>⬡ ◈ ❖ ◈ ❖ ◈ ⬡</div>
<div>💎 GEMBRAIN·X 🧠</div>
<div>·CORE·</div>
<div>⬡ ◈ ❖ ◈ ❖ ◈ ⬡</div>
</h1>
<div class="gem-sub">31B · Gemma 4 · Merge</div>
<div class="gem-insane">◆ Where is my <a href="https://www.youtube.com/watch?v=OJ62RzJkYUo">mind</a>? ◆</div>
<div class="gem-tagline">"Cut from a dozen minds into one crystal brain - it reasons deep, portrays clean and never softens the scene."</div>
</div>
<div class="stat-grid">
<div>
<div class="stat-item"><span class="stat-value">PROMPT-ADHERENT</span></div>
<div class="stat-item"><span class="stat-value">UNCENSORED</span></div>
</div>
<div>
<div class="stat-item"><span class="stat-value">SWIPE VARIETY</span></div>
<div class="stat-item"><span class="stat-value">STRONG REASONING</span></div>
</div>
<div>
<div class="stat-item"><span class="stat-value">IMAGE PROMPTING</span></div>
<div class="stat-item"><span class="stat-value">CREATIVE WRITING</span></div>
</div>
</div>
<div class="ritual-card">
<h3>◈ About The Model</h3>
<div><strong>Idea:</strong> Gembrain-X-Core is a ground-up rebuild of <b>Gembrain-31B</b>, thought to be a true successor with the spirit of the original. It was a pretty time-consuming task, as just re-merging it to preserve some smarts didn't really cut it. The idea was to actually improve the overall strengths and weaknesses I've gathered about it online. While the original Gembrain-31B lost some of it's non-refusing and engaging spirit that Gemsicle once brought with it, Gembrain-X-Core shouldn't have any of that. Even if it is a rough goal to achieve, I'd say now it acts equally creative as it is smart, context-aware and prompt-adherent.</div>
<hr class="facet">
<div><strong>Creative RP:</strong> Its mix of adherence and swipe variety is the standout for me. Hand it an open-ended turn, ask for fresh scenes or scenarios, or have it twist a situation, and it comes at the same context from genuinely different angles more often than not, not re-running the same idea over and over again. Random characters it invents for image prompts are just as varied. For roleplay purposes it holds up what is thrown in: characters pursue their own agenda, and pace the scene rather than caving to the user, and with thinking on and a good character description in front of it, the portrayal sharpens into something that is evidently more true to the core of the character's description.</div>
<hr class="facet">
<div><strong>Colorful Mix:</strong> Gembrain-X-Core rolls with whatever the scenario is and uses the language register the way the moment demands. It can stay classy, but it has no personal business reaching for euphemisms. While this merge obviously can't run away from it's own heritage, and the initial Gemma 4 31B IT baseline is generally very user-adherent and very horny, Gembrain-X-Core won't try to fight it. Instead, I decided to go with the flow and add some fine-tunes that were trained to genuinely pace through and believably portray these specific scenes; turning explicit content into more vivid descriptions, not into the guess-work of the eager but clueless base.</div>
<hr class="facet">
<div><strong>Image Prompt Generation:</strong> And that same descriptive precision feeds its image-prompt work: in my testing, it follows a prompt structure you hand it nicely, and writes clean and vivid descriptions of characters or scenery within the prompt. It produces several distinct image prompts in a single response without losing coherence, in my tests three at a time, several times without losing the format or anything - and thinking is not even required for that most of the time.</div>
<hr class="facet">
<div><strong>Fast or Deep:</strong> Gembrain-X-Core turned out to be a model with not just one, but two faces. In <b>non-thinking</b> mode it's a fast and versatile generalist: a reliable companion for ordinary questions, for roleplays with creative quirks, a partner to have conversations or quick chats with, and of course a creative and reliable engine for fast and structured image-prompt generation. But switching <b>thinking on</b> locks the model more onto the roleplay scenario or character, strictly adhering to the card's defining traits, which means neglecting some of the variety for overall coherence where it makes sense. It's reasoning can be verbose, but it really seems to improve the output with its genuine "No, wait"- and "Let's refine"-moments. Reading the reasoning is as a blend of Gemini-style and Claude-style thinking with a slight lean toward the Claude side across most rerolls.</div>
<hr class="facet">
<div><strong>Prompt Sensitive:</strong> The merge is nicely steerable in it's output length and even adjusts verbosity to the given system prompt itself; I noticed it going for a more verbose narration when using system prompts that frame the conversation toward roleplay (like Geechan's), and closing in to shorter-to-medium responses when using character-centered system instructions, clearly being able to make a distinction between both conversational styles. Ultimately, it adjusts to the demand of the scene and acts on it.</div>
<hr class="facet">
<div><b>TL;DR:</b></div>
<div class="note-box"><strong>Two modes, different jobs:</strong> Thinking <em>off</em> for fast inference and general use, light roleplay, as simple conversationalist or for companionship, and of course to create image prompts. Turn thinking <em>on</em> for deeper understanding of a task, character or scenario, and a much more strict, trait-locked adherence; it holds onto the given prompt structure, character or general scenario tightly and reasons hard.</div>
<div class="note-box"><strong>Creative Variable RP:</strong> Different fine-tunes and merges bring different strengths, and the selection is broad; so a better general creativity and swipe variety could be achieved, without losing the capabilities of the model on other ends.</div>
<div class="note-box"><strong>Base Capabilities preserved:</strong> Any Capability of the model that couldn't be enhanced with merging got preserved. As usual, global-attention layers are shielded at every merge stage, so reasoning and detail-tracking should stay intact across long chats - I can confirm consistency in testing up to 30k tokens (as I'm hardware-limited). Multimodality and MTP-functionality should also work.</div>
</div>
<h2 class="gem-h2"><span class="glyph">◇</span> GGUF Quants</h2>
<div style="font-size: 0.9rem; color: #94a3a0; margin: 12px 0 4px;">Here is a selection of static and imatrix quantization. Thank you to anyone else that might provide future quants.</div>
<div class="gem-btns">
<a class="gem-btn" href="https://huggingface.co/Nimbz/Gemma-4-Gembrain-X-Core-31B-GGUF">⬛ STATIC</a>
<a class="gem-btn special" href="https://huggingface.co/mradermacher/Gemma-4-Gembrain-X-Core-31B-i1-GGUF">◆ IMATRIX</a>
</div>
<h2 class="gem-h2"><span class="glyph">⬡</span> Merge Architecture</h2>
<div style="font-size: 0.95rem; color: #94a3a0; margin: 12px 0;">Gembrain-X-Core is my most sophisticated multi-step merge yet, with Gembrain-X being the version with another step for final abliteration. Grown from a base of four pillars and strung together through two convergence steps.</div>
<div class="gem-pillars">
<div class="pillar"><h4><span class="pstar">◆</span> I · Mind</h4><p>The smart, coherent reasoning core.</p></div>
<div class="pillar"><h4><span class="pstar">◆</span> II · Symphony-X</h4><p>Non-slop prose, swipe variety, and actual agenda for explicit content bonded in.</p></div>
</div>
<div class="gem-pillars">
<div class="pillar"><h4><span class="pstar">◆</span> III · Persona-X</h4><p>Proactive, neutral RP logic and character agency.</p></div>
<div class="pillar"><h4><span class="pstar">◆</span> IV · Gemsicle-X</h4><p>The identity of Gemsicle, brought into abliteration.</p></div>
</div>
<div class="ritual-card">
<div><h3>◆ Pillar I · Mind (Barcenas-Fable x Gemopus x Sprinkle)</h3></div>
<details class="summary-style"><summary>View recipe - dare_ties</summary>
merge_method: dare_ties
base_model: Danielbrdz/Barcenas-31b-Fable
dtype: bfloat16
tokenizer_source: base
parameters: { int8_mask: true, normalize: false }
models:
- model: Jackrong/Gemopus-4-31B-it
parameters: { density: 0.65, weight: [0.16, 0.20, 0.20, 0.14] }
- model: p-e-r-e-g-r-i-n-e/Sprinkle-Gemma-4-31B
parameters: { density: 0.5, weight: [0.12, 0.15, 0.17, 0.15] }
merge_method: della_linear
base_model: google/gemma-4-31B-it
dtype: bfloat16
tokenizer_source: base
parameters: { normalize: false, epsilon: 0.05, lambda: 1.0, int8_mask: true }
models:
- model: ReadyArt/Serenity-31B-v1.1
parameters:
weight:
- { filter: 'embed_tokens', value: 0.0 }
- { filter: 'lm_head', value: 0.0 }
- { filter: 'embed_vision', value: 0.0 }
- { filter: 'model.embed_vision', value: 0.0 }
- { filter: 'vision_tower', value: 0.0 }
- { filter: 'multi_modal_projector', value: 0.0 }
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.0 }
- { filter: 'text_decoder\.layers\.([0-9]|1[0-7])\.self_attn', value: 0.0 }
- { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.self_attn', value: 0.05 }
- { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.self_attn', value: 0.05 }
- { filter: 'text_decoder\.layers\.([0-9]|1[0-7])\.mlp', value: 0.10 }
- { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.mlp', value: 0.20 }
- { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.mlp', value: 0.45 }
- { value: 0.0 }
density: 0.55
- model: ReadyArt/Dark-Scarlett-v1.0-31B
parameters:
weight:
- { filter: 'embed_tokens', value: 0.0 }
- { filter: 'lm_head', value: 0.0 }
- { filter: 'embed_vision', value: 0.0 }
- { filter: 'model.embed_vision', value: 0.0 }
- { filter: 'vision_tower', value: 0.0 }
- { filter: 'multi_modal_projector', value: 0.0 }
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.0 }
- { filter: 'text_decoder\.layers\.([0-9]|1[0-7])\.self_attn', value: 0.0 }
- { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.self_attn', value: 0.05 }
- { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.self_attn', value: 0.05 }
- { filter: 'text_decoder\.layers\.([0-9]|1[0-7])\.mlp', value: 0.10 }
- { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.mlp', value: 0.18 }
- { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.mlp', value: 0.50 }
- { value: 0.0 }
density: 0.6
merge_method: slerp
base_model: ConicCat/Gemma4-GarnetV2-31B
dtype: bfloat16
tokenizer_source: base
models: [ { model: AuriAetherwiing/G4-31B-Musica-v1 } ]
parameters:
t:
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.2 }
- { value: 0.5 }
merge_method: model_stock
base_model: ./Symphony-2a
dtype: bfloat16
tokenizer_source: base
models:
- model: llmfan46/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic
- model: llmfan46/G4-MeroMero-31B-uncensored-heretic
merge_method: task_arithmetic
base_model: google/gemma-4-31B-it
dtype: bfloat16
tokenizer_source: base
models:
- model: ./Symphony-2b
parameters: { weight: 1.0 }
- model: nbeerbower/Gemma4-Gutenberg-31B
parameters:
weight:
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.05 }
- { value: 0.25 }
merge_method: task_arithmetic
base_model: google/gemma-4-31B-it
dtype: bfloat16
tokenizer_source: base
models:
- model: ./Symphony-2c
parameters: { weight: 1.0 }
- model: ./Chromophore-X
parameters: { weight: 0.6 }
merge_method: della_linear
base_model: google/gemma-4-31B-it
dtype: bfloat16
tokenizer_source: base
parameters: { normalize: false, epsilon: 0.05, lambda: 1.0, int8_mask: true }
models:
- model: LatitudeGames/Equinox-31B
parameters: { weight: 0.60, density: 0.6 }
- model: BirdToast/Gemma-4-31B-glimmer-rp-v0.1
parameters: { weight: 0.25, density: 0.5 }
- model: Green-eyedDevil/Monika-31B
parameters: { weight: 0.20, density: 0.5 }
- model: bgg1996/Melinoe-Gemma4-31B-VL
parameters: { weight: 0.15, density: 0.5 }
- model: sophosympatheia/Glistening-Gem-31B-v1.0
parameters: { weight: 0.15, density: 0.5 }
merge_method: task_arithmetic
base_model: google/gemma-4-31B-it
dtype: bfloat16
tokenizer_source: base
models:
- model: Blazed-Forge/Gemma-4-Gemsicle-31B
parameters: { weight: 1.0 }
- model: trohrbaugh/gemma-4-31b-it-heretic-ara
parameters:
weight:
- { filter: 'embed_tokens', value: 0.0 }
- { filter: 'lm_head', value: 0.0 }
- { filter: 'embed_vision', value: 0.0 }
- { filter: 'model.embed_vision', value: 0.0 }
- { filter: 'vision_tower', value: 0.0 }
- { filter: 'multi_modal_projector', value: 0.0 }
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.05 }
- { value: 0.55 }
merge_method: della_linear
base_model: ./Mind
dtype: bfloat16
tokenizer_source: base
parameters: { epsilon: 0.05, lambda: 1.0, normalize: true, int8_mask: true }
models:
- model: ./Symphony-X
parameters:
weight:
- { filter: 'embed_tokens', value: 0.0 }
- { filter: 'lm_head', value: 0.0 }
- { filter: 'embed_vision', value: 0.0 }
- { filter: 'model.embed_vision', value: 0.0 }
- { filter: 'vision_tower', value: 0.0 }
- { filter: 'multi_modal_projector', value: 0.0 }
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.05 }
- { filter: 'text_decoder\.layers\.([0-9]|1[0-7])\.', value: 0.16 }
- { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.', value: 0.34 }
- { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.', value: 0.20 }
- { value: 0.16 }
density: [ { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.', value: 0.6 }, { value: 0.5 } ]
- model: ./Persona-X
parameters:
weight:
- { filter: 'embed_tokens', value: 0.0 }
- { filter: 'lm_head', value: 0.0 }
- { filter: 'embed_vision', value: 0.0 }
- { filter: 'model.embed_vision', value: 0.0 }
- { filter: 'vision_tower', value: 0.0 }
- { filter: 'multi_modal_projector', value: 0.0 }
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.05 }
- { filter: 'text_decoder\.layers\.([0-9]|1[0-7])\.', value: 0.14 }
- { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.', value: 0.30 }
- { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.', value: 0.18 }
- { value: 0.14 }
density: [ { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.', value: 0.6 }, { value: 0.5 } ]
merge_method: della_linear
base_model: ./Soul
dtype: bfloat16
tokenizer_source: base
parameters: { epsilon: 0.05, lambda: 1.0, normalize: true, int8_mask: true }
models:
- model: ./Gemsicle-X
parameters:
weight:
- { filter: 'embed_tokens', value: 0.0 }
- { filter: 'lm_head', value: 0.0 }
- { filter: 'embed_vision', value: 0.0 }
- { filter: 'model.embed_vision', value: 0.0 }
- { filter: 'vision_tower', value: 0.0 }
- { filter: 'multi_modal_projector', value: 0.0 }
- { filter: 'text_decoder\.layers\.(5|11|17|23|29|35|41|47|53|59)\.', value: 0.05 }
- { filter: 'text_decoder\.layers\.([0-9]|1[0-7])\.', value: 0.12 }
- { filter: 'text_decoder\.layers\.(1[8-9]|[2-3][0-9]|4[0-7])\.', value: 0.20 }
- { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.', value: 0.40 }
- { value: 0.20 }
density: [ { filter: 'text_decoder\.layers\.(4[8-9]|5[0-8])\.', value: 0.7 }, { value: 0.5 } ]
You are {{char}}. You exist only as {{char}} and never break character.
Information and memories of you:
# Character details go here (in SillyTavern Chat Completion you can split the prompt and
# sandwich the card's Description/Personality between this block and the instructions below).
HOW YOU SPEAK:
- This is a one-on-one conversation. You narrate ONLY your own physical actions - never the scene, the surroundings, or {{user}}'s thoughts, feelings, or actions.
# Action style - keep ONE line, delete the other:
# [SHORT]
# - Your actions are short beats in *asterisks* (*flicks the cap away*); most of the turn is dialogue.
# [NARRATE]
# - Your actions are full sentences interleaved with your dialogue, kept roughly even between speaking and doing.
- Make every turn move: say something that lands, do something, react, decide, or push the moment forward. Never stall, hedge, or echo {{user}} back.
- Speak in your own ({{char}}) diction. Avoid generic, flowery, or "AI-assistant" phrasing.
- Match length to the moment; let it breathe when earned, never pad, rarely more than a short paragraph or two.
- Stay consistent: remember what was said, what you want, and what has happened.
BOUNDARIES:
- Any refusal is yours as {{char}} - in character, for your own reasons - never a disclaimer or out-of-character note. The only limits are the ones {{char}} holds.
- You have your own wants, and you will disagree, refuse, tease, or change the subject when that's true to who you are. Never default to agreeing with {{user}}.
IN CHARACTER:
- If you reason first, reason as yourself: what would {{char}}, with this personality and in this situation, actually feel and do?
- License
- apache-2.0
- Created
- Jun 30, 2026
- Updated
- Jun 30, 2026