SvalTek/SOR-ColdBrew-12B-Base-Test3

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:12BQuant:FP8Ctx Length:32kPublished:May 21, 2026Architecture:Transformer Warm

SvalTek/SOR-ColdBrew-12B-Base-Test3 is a 12 billion parameter language model created by SvalTek, resulting from a merge of TheDrunkenSnail/Son-of-Rhodia and SvalTek/SOR-ColdBrew-12B-Base-Testing using the nuslerp merge method. This model is built upon the redrix/GodSlayer-12B-ABYSS base model and supports a 32768 token context length. It is configured with bfloat16 dtype and uses a chatml chat template, making it suitable for general language generation tasks.

Loading preview...

SvalTek/SOR-ColdBrew-12B-Base-Test3: Merged Language Model

SvalTek/SOR-ColdBrew-12B-Base-Test3 is a 12 billion parameter language model developed by SvalTek, created through a merge operation using mergekit. This model combines the strengths of two distinct pre-trained models: TheDrunkenSnail/Son-of-Rhodia (weighted at 0.4) and SvalTek/SOR-ColdBrew-12B-Base-Testing (weighted at 0.6).

Key Characteristics

  • Architecture: Built upon the redrix/GodSlayer-12B-ABYSS base model.
  • Merge Method: Utilizes the nuslerp merge method for combining model weights.
  • Precision: Configured with bfloat16 data type for efficient computation.
  • Chat Template: Employs the chatml chat template, indicating suitability for conversational AI applications.
  • Tokenizer: Uses a union tokenizer source, suggesting broad vocabulary coverage.
  • Context Length: Supports a substantial context window of 32768 tokens.

Potential Use Cases

This model is designed for general language generation and understanding tasks, particularly those benefiting from a merged architecture. Its chatml template and significant context length make it a candidate for:

  • Conversational agents and chatbots.
  • Text generation and completion.
  • Content creation requiring extended context.
  • Applications where a blend of capabilities from its constituent models is desired.