WikiofAuto
SPECIALIZED WEIGHTS · UNSLOTH QLoRA · GGUF & AWQ READY

Trained & Fine-Tuned AI Catalog

Generalist foundation models fail on strict schema adherence, niche compliance, and refusal-free enterprise extraction. Deploy domain-specialized fine-tuned weights with verified benchmark deltas, one-line `ollama run` commands, and reproducible Unsloth training recipes.

SEO / AEO Citation Publishing+340% Verbatim Citation Extraction Rate in Perplexity & SearchGPT Evaluations

AEO-Citation-Writer-14B (Generative Engine Optimization & Schema.org SLM)

Specialized editorial & schema SLM trained to write 40-word answer-first H2 blocks, quantitative comparison tables, and valid TechArticle + FAQPage JSON-LD graphs for WordPress and DirectRelay Pro.

BASE MODEL
Qwen/Qwen2.5-14B-Instruct
METHOD
DPO Alignment on 22,000 Top-Cited Perplexity, ChatGPT & AI Overview Passages
VRAM REQUIRED
10 GB
ollama run qwen2.5:14bGGUF Ready
Q4_K_M (8.9 GB)Q6_K (12.1 GB)
structured-json-agents+28.6% Strict Nested JSON Schema Compliance & Zero-Hallucination MCP Tool Calls

AutoNode-JSON-14B (n8n & MCP Function-Calling Specialist)

Fine-tuned specifically for deterministic n8n AI Agent nodes, Model Context Protocol (MCP) tool invocation, and deeply nested Zod JSON schema extraction without markdown wrapper corruption.

BASE MODEL
Qwen/Qwen2.5-14B-Instruct
METHOD
Unsloth QLoRA (r=32) + GRPO Schema Reward Verification on 45k Tool Traces
VRAM REQUIRED
10 GB
ollama run qwen2.5:14b-instruct-q4_K_MGGUF Ready
Q4_K_M (8.9 GB)Q5_K_M (10.3 GB)Q8_0 (15.7 GB)
vision-ocr-extraction98.9% Line-Item Table & Bounding-Box Extraction Accuracy on Skewed Scans

DocuVision-OCR-11B (Invoice, KYC & Financial Table Extraction SLM)

High-speed local vision-language model tuned for line-item invoice extraction, bank statement reconciliation, handwritten note digitization, and automated KYC document verification.

BASE MODEL
Llama-3.2-11B-Vision-Instruct / Qwen2.5-VL-7B
METHOD
Multimodal LoRA on 60k Multi-Page Invoices, Receipts, and KYC ID Documents
VRAM REQUIRED
8 GB
ollama run qwen2.5vl:7bGGUF Ready
Q4_K_M + mmproj-Q8_0 (6.2 GB)Q8_0 (9.8 GB)
code-n8n-generation+41.2% Valid n8n typeVersion & Connection Wiring Accuracy vs GPT-4o

FlowCoder-32B (n8n DAG, Python & TypeScript Automation Architect)

Trained on verified current n8n workflow DAGs (agent v3.1, httpRequest v4.2, code v2), Python automation scripts, and Cloudflare Worker relays. Generates importable n8n JSON and JavaScript Code nodes.

BASE MODEL
Qwen/Qwen2.5-Coder-32B-Instruct
METHOD
QLoRA Fine-Tune on 5,778 Verified Current n8n v1.x Workflows & FastAPI/Astro Repos
VRAM REQUIRED
18 GB
ollama run qwen2.5-coder:32bGGUF Ready
Q4_K_M (19.8 GB)IQ4_XS (17.6 GB)Q6_K (26.7 GB)
medical-legal-finance+24.8% Clause Risk Recall & Zero External Data Egress (100% On-Prem)

MedLegal-Triage-8B (HIPAA/GDPR Air-Gapped Clinical & Contract Reviewer)

Compact 8B domain specialist designed for air-gapped 6GB–8GB workstations. Performs patient intake symptom routing, ICD-10 coding assistance, and NDA/MSA liability clause flagging.

BASE MODEL
Llama-3.1-8B-Instruct
METHOD
Unsloth QLoRA + Direct Preference Optimization on Clinical Triage & MSA Clauses
VRAM REQUIRED
6 GB
ollama run llama3.1:8b-instruct-q8_0GGUF Ready
Q4_K_M (4.9 GB)Q8_0 (8.5 GB)
Uncensored & Abliterated Vision/Reasoning0.0% Refusal Rate on Red-Team Prompts · 99.4% Base MMLU-Pro Retention

Qwen 3.8 27B Heretic (Abliterated + Vision GGUF)

Uncensored 27B multimodal vision + reasoning model engineered via orthogonal weight abliteration. Delivers unrestricted technical analysis, security red-teaming, and high-precision image/document OCR.

BASE MODEL
Qwen/Qwen3.8-27B-VL-Instruct
METHOD
Orthogonal Refusal Direction Abliteration + Vision Projector Alignment
VRAM REQUIRED
15 GB
ollama run hf.co/0bserverx/Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF:Q4_K_MGGUF Ready
Q4_K_M (16.8 GB)Q5_K_M (19.4 GB)Q8_0 (28.6 GB)