czocelot/gemma4_12b_fable5lora_merged

TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 8, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The czocelot/gemma4_12b_fable5lora_merged model is a 12 billion parameter language model based on the Gemma4 architecture, fine-tuned from gemma4-12B-it. This model incorporates a LoRA (Low-Rank Adaptation) that was trained on a specific dataset and subsequently merged into the base model. It is designed for general language tasks, offering both BF16 precision and quantized versions for deployment flexibility. Its development focuses on leveraging the Gemma4 base with specialized adaptation.

Loading preview...

Model Overview

The czocelot/gemma4_12b_fable5lora_merged is a 12 billion parameter language model, building upon the gemma4-12B-it base architecture. This model has undergone a specific fine-tuning process involving a LoRA (Low-Rank Adaptation) module. The LoRA was trained on a distinct dataset and then merged directly into the foundational gemma4-12B-it model, enhancing its capabilities for various language-related tasks.

Key Characteristics

  • Base Model: Derived from the gemma4-12B-it model, indicating a strong foundation in instruction-tuned language understanding.
  • Parameter Count: Features 12 billion parameters, placing it in the medium-to-large scale LLM category.
  • Context Length: Supports a context window of 32768 tokens, allowing for processing and generating longer sequences of text.
  • Fine-tuning Method: Utilizes LoRA for efficient adaptation, where a low-rank adaptation was trained on a specific dataset and then merged.
  • Availability: Provided in both BF16 (Brain Floating Point 16-bit) precision and quantized versions, offering options for performance and memory efficiency.

Potential Use Cases

This model is suitable for applications requiring a robust language model with specialized fine-tuning. Its merged LoRA suggests potential strengths in areas related to the dataset it was trained on, making it a candidate for:

  • General text generation and completion.
  • Instruction following and conversational AI.
  • Tasks benefiting from a model adapted to specific data distributions.
  • Applications where both high precision and quantized versions are beneficial for deployment.

Finetunes
1 models