simpledirect/Vinci-Piccolo-1.0

VISIONConcurrency Cost:1Model Size:4.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jun 24, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Vinci Piccolo 1.0 by SimpleDirect is a 4.5 billion parameter, open-weight chat model fine-tuned from Qwen/Qwen3.5-4B with a 262,144 token context length. It is uniquely optimized for a consistent character and honest disposition, adhering to a public Constitution that defines its behavior. This model excels in conversational tasks, everyday questions, and drafting where character and honesty are prioritized over frontier reasoning capabilities, and is small enough for local or on-device inference.

Loading preview...

Vinci Piccolo 1.0: Character-Driven Chat Model

Vinci Piccolo 1.0, developed by SimpleDirect, is a 4.5 billion parameter, open-weight chat model built upon Qwen/Qwen3.5-4B. Its primary differentiator is a unique fine-tuning approach focused on instilling a consistent character and honest disposition, guided by a written, public Constitution. This model prioritizes ethical behavior and transparency, designed to abstain from fabricating answers when uncertain.

Key Capabilities & Characteristics

  • Character and Honesty: Fine-tuned to maintain a consistent personality and an honest disposition, preferring to abstain rather than invent information.
  • Constitutional Fine-Tuning: Behavior is governed by a public 'Constitution' that dictates its conversational style, values, and how it handles uncertainty.
  • Compact Size: At 4.5 billion parameters, it is designed to be small enough for efficient local and on-device inference.
  • Extended Context: Features a substantial 262,144 token context window.
  • Safety & Robustness: Demonstrates strong adversarial robustness, with low attack success rates on HarmBench (2.5%) and JailbreakBench (1.0%).

Intended Use Cases

  • Conversational AI: Ideal for general conversation, answering everyday questions, and drafting where a pleasant and honest interaction is key.
  • Local & On-Device Applications: Suitable for deployment on laptops and other edge devices due to its modest size.
  • Base for Experimentation: Can serve as a foundation for further fine-tuning or research into character-driven AI.

Limitations

While strong in character and honesty, Vinci Piccolo 1.0 is not designed for complex tasks requiring frontier reasoning, advanced mathematics, or intricate coding. Its tool/function calling capabilities are limited at this size, and users should verify any critical information as it is not a source of ground truth. French language support is partial, lagging behind English.

Finetunes
2 models