1,797 Downloads Updated 1 month ago
The JOSIE models are a family of uncensored, high-performance language models developed by Gökdeniz Gülmez that blend a friendly, human-like personality with specialized capabilities for natural conversation and complex analytical reasoning.
ollama run goekdenizguelmez/JOSIE
Applications
ollama launch claude --model goekdenizguelmez/JOSIE
ollama launch codex --model goekdenizguelmez/JOSIE
ollama launch opencode --model goekdenizguelmez/JOSIE
ollama launch openclaw --model goekdenizguelmez/JOSIE
Models
View all →Readme
J.O.S.I.E. Model Family
Just One Super Intelligent Entity — A production-grade AI family developed by Gökdeniz Gülmez, engineered for absolute freedom, human-like personality, and high-performance reasoning on consumer hardware.
🌟 Model Overview
The JOSIE family represents a paradigm shift in open-source AI, moving away from robotic sycophancy toward genuine, direct, and helpful interaction. All models are Gabliterated (uncensored) at the base level to ensure they provide real answers without artificial constraints.
| Feature | ⚡ JOSIE-Instruct | 🧠 JOSIE-Thinking |
|---|---|---|
| Primary Focus | Daily Chat & Assistance | Complex Logic & STEM |
| Response Style | Natural, Human-like, Emoji-rich | Analytical, Structured CoT |
| Context Window | 32,768 Tokens | 65,536 Tokens |
| Personality | Friendly, expressive, honest | Direct, analytical, objective |
| Best Utility | Coding, Emails, Creative Chat | Math, Logic, Large Docs |
| Base Architecture | Qwen3-4B-Instruct (Gabliterated) | Qwen3-4B-Thinking (Gabliterated) |
🧬 The Two Pillars of JOSIE
1. ⚡ JOSIE-Instruct (The Conversationalist)
Designed to be your everyday companion. It doesn’t just “talk”; it interacts. - Human-Centric: Uses emojis 🐍✨ and natural phrasing to break the “AI wall.” - Direct & Honest: No “as an AI model” lecturing. Just answers. - Versatile: Master of creative writing, quick coding fixes, and roleplay.
2. 🧠 JOSIE-Thinking (The Architect)
When the problem is deep, JOSIE-Thinking goes deeper. - Chain-of-Thought: Automatically breaks down complex tasks into logical steps. - STEM Specialized: High proficiency in Calculus, Physics, and Data Science. - Heavy Lifting: Distilled from high-tier reasoning traces for maximum cognitive output.
🚀 Technical Excellence
- Framework: Built using
mlx-lm-lora. - Method: Full-weight fine-tuning for deep knowledge integration.
- No Filters: Designed for professionals who require unfiltered, high-fidelity responses.