kairawal/Gemma-3-4B-IT-EN-SynthDolly-r16alpha128-E5-S3407

Hugging Face
VISIONConcurrency Cost:1Model Size:4.3BQuant:BF16Ctx Length:32kPublished:May 24, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The kairawal/Gemma-3-4B-IT-EN-SynthDolly-r16alpha128-E5-S3407 is a 4.3 billion parameter instruction-tuned Gemma model, finetuned by kairawal. This model was optimized for training speed using Unsloth and Huggingface's TRL library, offering efficient performance for English language tasks. With a 32768 token context length, it is suitable for applications requiring processing of longer inputs and generating coherent, instruction-following responses.

Loading preview...

Model Overview

The kairawal/Gemma-3-4B-IT-EN-SynthDolly-r16alpha128-E5-S3407 is an instruction-tuned Gemma model with approximately 4.3 billion parameters, developed by kairawal. It is finetuned from the unsloth/gemma-3-4b-it base model and utilizes a substantial 32768 token context window, making it capable of handling extensive conversational or document-based inputs.

Key Characteristics

  • Architecture: Based on the Gemma 3.4B instruction-tuned architecture.
  • Parameter Count: 4.3 billion parameters, offering a balance between performance and computational efficiency.
  • Context Length: Supports a 32768 token context, enabling processing of long texts and complex instructions.
  • Training Optimization: Finetuned using Unsloth and Huggingface's TRL library, which significantly accelerated the training process.
  • Language: Primarily designed for English language instruction-following tasks.

Use Cases

This model is well-suited for applications requiring:

  • Instruction Following: Generating responses based on explicit instructions.
  • Long Context Processing: Tasks that involve understanding and summarizing lengthy documents or conversations.
  • Efficient Deployment: Its optimized training suggests potential for efficient inference, making it suitable for resource-constrained environments where a 4.3B parameter model is appropriate.