Romance Novel AI Command Center

Master Model Card Reference • Optimized for NovelCrafter Sliders (14 GB RAM Budget & 8-bit KV-Cache)

Qwenvergence 14B v12 Prose-DS

hf.co/.../Qwenvergence-14B-v12-Prose-DS-GGUF:Q5_K_M
Tier 1
Parameters14B
Size on Disk~10.5 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
2.1 t/s
5070 Ti GPU
55 t/s
Full Description

Elite-tier deep prose intelligence merge. Tailored explicitly for sophisticated narrative styling, atmospheric pacing, and rich literary depth.

NovelCrafter Application

Ideal as your primary author model for complex emotional arcs, scene drafting, and high-stakes romantic tension.

Medius Erebus Magnum 14B

hf.co/.../medius-erebus-magnum-14b-GGUF:Q5_K_M
Tier 1
Parameters14B
Size on Disk~10.5 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
2.1 t/s
5070 Ti GPU
58 t/s
Full Description

Engineered to emulate high-end proprietary creative writing models. Exceptional dynamic scene pacing, raw emotional weight, and zero prudish filters.

NovelCrafter Application

Incredible for writing intimate physical sequences, dramatic confrontations, and deep interior character monologues.

Magnum 12b v2.5 KTO

hf.co/.../magnum-12b-v2.5-kto-GGUF:Q5_K_M
Tier 1
Parameters12B
Size on Disk~8.8 GB
Max Context65,536 tokens
Pages Equiv.~260 Pages
i7 6700 CPU
2.6 t/s
5070 Ti GPU
68 t/s
Full Description

Trained using KTO preference alignment on top of Mistral Nemo architecture to replicate premier conversational and narrative prose flow.

NovelCrafter Application

Fantastic for multi-chapter continuity, maintaining complex plot timelines, and natural, witty dialogue banter.

MN-12B-Mag-Mell-R1

hf.co/.../MN-12B-Mag-Mell-R1-GGUF:Q5_K_M
Tier 1
Parameters12B
Size on Disk~8.8 GB
Max Context65,536 tokens
Pages Equiv.~260 Pages
i7 6700 CPU
2.6 t/s
5070 Ti GPU
68 t/s
Full Description

Multi-stage hyper-merge combining specialized narrative tropes, world-building grounding, and fluid prose. (Note: "R1" denotes revision level, not DeepSeek reasoning).

NovelCrafter Application

Use for balancing intricate romance plots with vivid setting descriptions and stable character reactions.

Huihui Gemma-4 12B Abliterated

hf.co/.../Huihui-gemma-4-12B-it-abliterated-GGUF:Q5_K_M
Tier 1
Parameters12B
Size on Disk~9.2 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
2.5 t/s
5070 Ti GPU
65 t/s
Full Description

Cutting-edge architecture with removal of artificial safety refusals, retaining maximum structural instruction adherence and rich creative vocabulary.

NovelCrafter Application

Great for enforcing strict formatting rules while writing deep, unrestricted character interactions.

Gemma 3 12B Instruct Abliterated

hf.co/.../gemma-3-12b-it-abliterated-refined-novis-GGUF:Q5_K_M
Tier 2
Parameters12B
Size on Disk~9.3 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
2.5 t/s
5070 Ti GPU
65 t/s
Full Description

Stripped of built-in refusal filters and specialized for rich semantic output without visual encoder weight bloat.

NovelCrafter Application

Exceptional for driving creative suggestions and handling heavy historical or contemporary romance subplots.

Qwen2.5 14B Instruct Abliterated v2

hf.co/.../Qwen2.5-14B-Instruct-abliterated-v2-GGUF:Q5_K_M
Tier 2
Parameters14B
Size on Disk~10.4 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
2.2 t/s
5070 Ti GPU
56 t/s
Full Description

Combines Qwen 2.5's immense multilingual and logical robustness with structural un-censoring for unrestricted creative freedom.

NovelCrafter Application

Strong option when your romance novel incorporates complex technical background details, legal plots, or medical drama.

Richardyoung Qwen2.5 14B

richardyoung/qwen2.5-14b-instruct-abliterated:latest
Tier 2
Parameters14B
Size on Disk~10.4 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
2.2 t/s
5070 Ti GPU
56 t/s
Full Description

Alternative native distribution of the 14B abliterated Qwen engine, offering stable local parsing and rigid formatting.

NovelCrafter Application

Use for generating structured chapter outlines, character beat sheets, and plot consistency checks.

Shisa-v2 Mistral Nemo 12B

hf.co/.../Shisa-v2-Mistral-Nemo-12B-Abliterated-GGUF:Q5_K_M
Tier 2
Parameters12B
Size on Disk~8.8 GB
Max Context65,536 tokens
Pages Equiv.~260 Pages
i7 6700 CPU
2.6 t/s
5070 Ti GPU
68 t/s
Full Description

Built on Mistral Nemo, fine-tuned for high nuance, excellent cross-lingual handling, and complete lack of preachy refusal behavior.

NovelCrafter Application

Great for international settings or multi-perspective romance narratives requiring natural voice switching.

Fimbulvetr 11B v2

hf.co/.../Fimbulvetr-11B-v2-GGUF:Q5_K_M
Tier 2
Parameters11B
Size on Disk~7.9 GB
Max Context65,536 tokens
Pages Equiv.~260 Pages
i7 6700 CPU
3.1 t/s
5070 Ti GPU
75 t/s
Full Description

Legendary community-loved Solar architecture fine-tune built explicitly for long-form creative writing and immersive roleplay.

NovelCrafter Application

An old favorite for romance authors; handles romantic tension and physical pacing with organic prose flow.

Mistral Nemo Instruct Abliterated

hf.co/.../Mistral-Nemo-Instruct-2407-abliterated-i1-GGUF:Q4_K_M
Tier 2
Parameters12B
Size on Disk~7.2 GB
Max Context128,000 tokens
Pages Equiv.~510 Pages
i7 6700 CPU
3.4 t/s
5070 Ti GPU
82 t/s
Full Description

Abliterated iteration of the massive-context Mistral Nemo engine, allowing up to 128k native context with a lightweight 4-bit footprint.

NovelCrafter Application

Perfect for dropping entire manuscripts or multi-act outlines into NovelCrafter for broad structural context tracking.

Gemma 2 9B Instruct Abliterated

hf.co/QuantFactory/.../gemma-2-9b-it-abliterated-GGUF:Q5_K_M
Tier 2
Parameters9B
Size on Disk~6.6 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
3.8 t/s
5070 Ti GPU
92 t/s
Full Description

Google's highly praised 9B model treated with abliteration. Offers exceptional emotional intelligence and poetic flair in dialogue.

NovelCrafter Application

Use for writing sparkling banter, emotional heart-to-hearts, and intimate romantic exchanges.

DeepSeek-R1 8B

deepseek-r1:8b
Tier 3
Parameters8B
Size on Disk~4.9 GB
Max Context65,536 tokens
Pages Equiv.~260 Pages
i7 6700 CPU
4.5 t/s
5070 Ti GPU
105 t/s
Full Description

Reasoning model optimized for internal chain-of-thought logic. Excellent at deduction and structural tracking, though prone to verbose thought tags.

NovelCrafter Application

Use behind the scenes in NovelCrafter for plotting out mystery subplots, character motivations, and timeline alignment.

Qwen 2.5 7B Instruct

qwen2.5:7b
Tier 3
Parameters7B
Size on Disk~4.7 GB
Max Context128,000 tokens
Pages Equiv.~510 Pages
i7 6700 CPU
5.2 t/s
5070 Ti GPU
115 t/s
Full Description

Extremely fast, highly obedient general instruction model with enormous native context and sharp reasoning.

NovelCrafter Application

Great as a fast utility workhorse for summarizing lorebook entries or structuring chapter summaries.

Llama 3.1 8B Abliterated

mannix/llama3.1-8b-abliterated:latest
Tier 3
Parameters8B
Size on Disk~4.8 GB
Max Context128,000 tokens
Pages Equiv.~510 Pages
i7 6700 CPU
4.6 t/s
5070 Ti GPU
108 t/s
Full Description

Uncensored community version of Meta's foundational Llama 3.1 8B architecture. Rock-solid instruction following.

NovelCrafter Application

Reliable general-purpose writer for standard romance tropes and broad scene generation.

Llama 3.1 8B

llama3.1:latest
Tier 3
Parameters8B
Size on Disk~4.8 GB
Max Context128,000 tokens
Pages Equiv.~510 Pages
i7 6700 CPU
4.6 t/s
5070 Ti GPU
108 t/s
Full Description

Meta's standard baseline instruct model. Highly stable structure, though subject to standard safety guardrails.

NovelCrafter Application

Good for general brainstorming where policy restrictions won't interfere with your romance plot.

Medius Erebus Magnum 14B (Q4)

hf.co/.../medius-erebus-magnum-14b-GGUF:Q4_K_M
Tier 3
Parameters14B
Size on Disk~8.7 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
2.7 t/s
5070 Ti GPU
68 t/s
Full Description

Slightly lighter 4-bit quantized variant of the elite Magnum 14B creative prose model, preserving most of its style.

NovelCrafter Application

Fallback prose model if you need extra RAM overhead on your mini PC for other concurrent apps.

Qwen2.5-VL 7B

qwen2.5vl:7b
Tier 4
Parameters7B + Vision
Size on Disk~6.0 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
4.0 t/s
5070 Ti GPU
95 t/s
Full Description

Multimodal Vision Powerhouse. Dedicated image encoder capable of granular visual breakdowns (fine physical traits, clothing, lighting, scars).

NovelCrafter Application

Photo-to-Prose: Feed actor faceclaims or mood board photos into this model to generate rich, hyper-detailed character descriptions for your NovelCrafter lorebook.

Llama 3.2 Vision 11B

llama3.2-vision:latest
Tier 4
Parameters11B + Vision
Size on Disk~7.8 GB
Max Context16,384 tokens
Pages Equiv.~65 Pages
i7 6700 CPU
3.2 t/s
5070 Ti GPU
76 t/s
Full Description

Meta's native multimodal vision model. Fast multi-image analysis and cross-modal reasoning capabilities.

NovelCrafter Application

Use to inspect environmental concept art or setting photos, translating visual details straight into descriptive prose snippets.

Phi-4 Mini

phi4-mini:latest
Tier 4
Parameters3.8B
Size on Disk~2.5 GB
Max Context128,000 tokens
Pages Equiv.~510 Pages
i7 6700 CPU
8.5 t/s
5070 Ti GPU
160 t/s
Full Description

Microsoft's ultra-compact, highly sophisticated small language model with surprisingly strong reasoning and synthetic data capabilities.

NovelCrafter Application

Great for quick background tasks, naming characters, or generating rapid-fire scene prompt inspirations.

Llama 3.2 3B

llama3.2:latest
Tier 4
Parameters3B
Size on Disk~2.0 GB
Max Context128,000 tokens
Pages Equiv.~510 Pages
i7 6700 CPU
11.0 t/s
5070 Ti GPU
210 t/s
Full Description

Lightweight edge model designed for rapid token generation and low RAM overhead.

NovelCrafter Application

Use for instant auto-complete suggestions or lightweight tag formatting in NovelCrafter.

DeepSeek-R1 1.5B

deepseek-r1:1.5b
Tier 4
Parameters1.5B
Size on Disk~1.1 GB
Max Context32,768 tokens
Pages Equiv.~130 Pages
i7 6700 CPU
18.5 t/s
5070 Ti GPU
320 t/s
Full Description

Tiny reasoning distillation. Extremely fast, though limited in complex prose context and long-form narrative consistency.

NovelCrafter Application

Useful only for quick structural logic checks where deep prose depth isn't required.