JOSIE-1.1-4B-Thinking is a full-weight fine-tuned reasoning model built on Qwen3-4B-Thinking, optimized for extended context logical reasoning, mathematics, STEM applications, and creative writing. The model was trained on a curated distillation dataset combining: 1. Reasoning Traces: Distilled from Josie-Zero-8B reasoning outputs 2. Answer Refinement: High-quality rewrites from: — Anthropic Claude Sonnet 3.7 — Anthropic Claude Sonnet 4.0 — Anthropic Claude Opus 4.5 — Anthropic Claude Opus 4.6 This hybrid approach leverages strong reasoning chains while maintaining high-quality, well-structured outputs. — Training Framework: MLX-LM-LoRA — Hardware: Apple Silicon (M-series chips) — Fine-tuning Method: Full-weight fine-tuning — Total Training Tokens: 1B+ — Training Repository: Custom MLX-LM-LoRA implementation (available on GitHub) The model was fine-tuned with the following system instruction: This system prompt establishes the model’s identity and capability framework, emphasizing maximum intelligence and capability without artificial restrictions. The model was trained exclusively on Apple Silicon using optimized MLX frameworks, demonstrating the viability of high-quality model…
Модальности:
Генерация текста
Области применения:
Диалог / чат Мультиязычность
Задача: Генерация текста
Автор: Goekdeniz-Guelmez
Теги: qwen3, chat, conversational, multilingual, en, es, fr, pt
Лайков: 4 | Загрузок: 31
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
