CalmState/gemma-3-4b-polyglot-v1

CalmState/gemma-3-4b-polyglot-v1: Specialized Gemma 3 Fine-Tune

This model is a specialized fine-tune of Google's Gemma 3 4B developed by CalmState, featuring 4.3 billion parameters and a 32768 token context length. It is specifically optimized for the Polyglot Air desktop application workflow, focusing on strict adherence to suffix-based commands. Unlike general-purpose chat models, this fine-tune is engineered to provide instant, clean, and direct responses without conversational filler or preambles, treating ::suffix as a strict instruction code.

Key Capabilities

• Strict Instruction Following: Designed to process suffix-based commands for precise transformations.
• Multilingual Translation: Supports translation to English (::en), Portuguese (Portugal ::pt, Brazil ::ptbr), Spanish (::es), and Chinese (::zh).
• Text Transformation: Includes functions for grammar and spelling correction (::fix), rewriting in a formal tone (::formal), and rewriting in a casual tone (::casual).
• Summarization: Capable of generating concise summaries of longer texts (::summary).
• Optimized for Integration: Built for seamless use within the Polyglot Air application, ensuring direct and non-conversational output.

Good For

• Developers and users requiring a model that strictly follows commands for text manipulation.
• Applications needing direct, non-conversational outputs for tasks like translation or text rephrasing.
• Integration into desktop applications like Polyglot Air for efficient, instruction-driven text processing.