bigatuna/Qwen3.5-9b-Sushi-Coder
bigatuna/Qwen3.5-9b-Sushi-Coder is a 9 billion parameter language model, fine-tuned from unsloth/qwen3.5-9b. This model specializes in reasoning tasks, having undergone continuation training on the nohurry/Opus-4.6-Reasoning-3000x-filtered dataset. It is optimized for code-related applications and complex reasoning, leveraging its 32768 token context length.
Loading preview...
Model Overview
bigatuna/Qwen3.5-9b-Sushi-Coder is a 9 billion parameter language model, fine-tuned by bigatuna from the unsloth/qwen3.5-9b base model. It has been specifically developed for enhanced reasoning capabilities, building upon an earlier training lineage that included open-r1/codeforces-cots.
Key Characteristics
- Base Model: Derived from
unsloth/qwen3.5-9b. - Specialized Training: Underwent continuation training using the
nohurry/Opus-4.6-Reasoning-3000x-filtereddataset, focusing on reasoning tasks. - Training Method: Utilizes LoRA continuation from an adapter-only Unsloth Studio output, with 16-bit LoRA and bf16 precision.
- Context Length: Features a 32768 token context window, suitable for handling extensive inputs.
- Development Tools: Built with Unsloth and TRL, ensuring efficient fine-tuning and deployment.
Ideal Use Cases
This model is particularly well-suited for applications requiring strong reasoning abilities and code-related tasks, given its specialized training data. Its large context window also makes it effective for processing and generating longer sequences of text or code.