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

TEXT GENERATIONPricing:Input $2.72 / Output $4.8Concurrent Unit Cost:2Model Size:32.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Feb 10, 2025License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

nitky/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-ABEJA-32B-Preview is a 32.8 billion parameter language model merged using the SCE method, based on abeja/ABEJA-Qwen2.5-32b-Japanese-v0.1. It integrates components from nitky/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-32B-Preview, leveraging a Qwen2.5-32B architecture with a 32768 token context length. This model is specifically designed for Japanese language tasks, demonstrating enhanced reasoning capabilities, particularly in code generation scenarios where it can produce detailed thought processes and contextually appropriate responses.

Loading preview...

Model Overview

nitky/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-ABEJA-32B-Preview is a 32.8 billion parameter language model created by nitky through a merge of several pre-trained models using the mergekit tool. This model utilizes the SCE (Sub-Component Ensemble) merge method, building upon abeja/ABEJA-Qwen2.5-32b-Japanese-v0.1 as its base. It integrates nitky/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-Japanese-32B-Preview to enhance its capabilities.

Key Capabilities

  • Advanced Code Generation: Demonstrated ability to generate Python code, including complex tasks like FizzBuzz, with detailed reasoning steps and context-aware explanations.
  • Role-Playing and Contextual Understanding: Excels at adopting specific personas (e.g., Oda Nobunaga teaching Hideyoshi) and integrating thematic elements into its responses, including code comments and explanations.
  • Japanese Language Proficiency: Optimized for Japanese language tasks, making it suitable for applications requiring nuanced understanding and generation in Japanese.
  • Reasoning Token Output: Capable of producing <think> tokens that outline its thought process before generating a final response, offering transparency into its reasoning.

Good for

  • Japanese Code Generation: Developers needing a robust model for generating code in Python with detailed, context-rich explanations in Japanese.
  • Creative Content Generation: Scenarios requiring the model to adopt specific personas or integrate narrative elements into its output.
  • Educational Tools: Creating interactive learning experiences where explanations and reasoning are crucial.
  • Complex Instruction Following: Tasks that involve multi-step instructions and require the model to break down problems and explain its approach.

Finetunes
1 models