nitky/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-Bakeneko-32B-Preview

TEXT GENERATIONPricing:Input $2.72 / Output $4.8Concurrent Unit Cost:2Model Size:32.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Feb 13, 2025License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

nitky/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-Bakeneko-32B-Preview is a 32.8 billion parameter language model created by nitky, merged using the SCE method with rinna/qwen2.5-bakeneko-32b as a base. This model is specifically designed for Japanese language tasks, demonstrating advanced reasoning capabilities as shown in its detailed thought process for code generation. Its primary use case is complex Japanese instruction following and code generation, particularly when nuanced, context-aware responses are required.

Loading preview...

FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-Bakeneko-32B-Preview Overview

This model, developed by nitky, is a 32.8 billion parameter language model optimized for Japanese language processing. It was created using the SCE (Supervised Contrastive Embedding) merge method, combining rinna/qwen2.5-bakeneko-32b as a base with nitky/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-32B-Preview. The model demonstrates a notable ability for complex instruction following and detailed reasoning, particularly in code generation tasks.

Key Capabilities

  • Advanced Japanese Instruction Following: Excels at understanding and executing intricate instructions provided in Japanese.
  • Detailed Reasoning: Showcases a robust internal thought process (<think> tags) to break down complex problems, as exemplified in its FizzBuzz code generation with a unique "Oda Nobunaga" persona.
  • Code Generation: Capable of generating functional Python code based on natural language prompts, including creative and contextual constraints.
  • Merged Architecture: Leverages the strengths of its constituent models through the SCE merge method, enhancing its overall performance for Japanese tasks.

Good For

  • Japanese-centric Applications: Ideal for developers building applications that require high-quality Japanese text generation and understanding.
  • Complex Problem Solving: Suitable for tasks demanding detailed logical reasoning and step-by-step problem decomposition.
  • Creative Code Generation: Useful for scenarios where code needs to be generated with specific stylistic or contextual requirements, such as role-playing personas.
  • Research and Development: Provides a strong foundation for further fine-tuning or experimentation in Japanese LLM capabilities.

Finetunes
1 models