PersianML/gemma-2-2b-persian-v2

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2.6BQuant:BF16Context Size:8kPublished:Jul 22, 2026License:gemmaArchitecture:Transformer0.0K Featherless Exclusive Cold

PersianML/gemma-2-2b-persian-v2 is an optimized 2.6 billion parameter Gemma-2-2b-it based model, specifically enhanced for Persian language conversational tasks. Developed by PersianML, this experimental version incorporates self-merging techniques to improve performance and efficiency in generating Persian text. It builds upon Google's Gemma-2-2b-it and previous fine-tuning efforts, focusing on research and community development for Persian LLMs.

Loading preview...

Overview

PersianML/gemma-2-2b-persian-v2 is an experimental, optimized version of the Gemma-2-2b-it model, specifically fine-tuned for the Persian language. Building on Google's Gemma-2-2b-it and mshojaei77/Gemma-2b-fa, this 2.6 billion parameter model incorporates self-merging and other experimental techniques to enhance its performance and efficiency in Persian conversational tasks.

Key Capabilities & Features

  • Optimized for Persian: Enhanced for generating Persian text and engaging in conversations.
  • Self-Merged Architecture: Utilizes self-merging for a potentially more robust and coherent representation.
  • Gemma-2-2b-it Foundation: Based on Google's Gemma-2-2b-it, ensuring a strong architectural base.
  • Experimental Development: Under active development, focusing on research and community contributions.

Intended Use Cases

  • Research and Experimentation: Ideal for exploring self-merging techniques on Persian Gemma models.
  • Educational Purposes: Useful for demonstrating advanced fine-tuning and optimization methods.
  • Community Development: Contributes to the growing ecosystem of Persian language models.
  • Prototyping: Suitable for early-stage prototyping, with an understanding of its experimental nature and variable output quality.

Limitations

As an experimental 2.6 billion parameter model, it has limitations including variable output quality, potential for imperfections like fluency issues or factual inaccuracies, and performance constraints compared to larger models. Users should exercise caution and critical evaluation.

Finetunes
1 models