espressovi/BODHI-gemma-3-12b-distil
espressovi/BODHI-gemma-3-12b-distil is a math-focused language model created by espressovi through Long-CoT distillation from Google's Gemma-3-12b-pt base model. This model is specifically optimized for mathematical reasoning tasks, leveraging its distillation process to enhance performance in this domain. It is designed to provide strong capabilities in solving complex mathematical problems.
Loading preview...
Model Overview
The espressovi/BODHI-gemma-3-12b-distil is a specialized language model developed as part of the BODHI project. It is derived from google/gemma-3-12b-pt through a technique called Long-CoT distillation, which aims to enhance its capabilities in specific areas.
Key Capabilities
- Mathematical Reasoning: The primary focus of this model is on mathematical tasks, indicating a strong performance in problem-solving within this domain.
- Distillation Process: Utilizes Long-CoT distillation, suggesting an optimized learning process for complex reasoning chains.
Performance
Initial performance metrics are provided for AIME25 at T=1.0, 16K tokens, and pass@8, indicating its evaluation against challenging mathematical benchmarks. While specific scores are not detailed in this summary, the mention of AIME25 highlights its intended application in advanced mathematics.