T

Tinyr1 32B Preview

Developed by qihoo360
Tiny-R1-32B-Preview is an inference model based on Deepseek-R1-Distill-Qwen-32B, focusing on mathematics, coding, and scientific fields, with performance close to the full version of the R1 model.
Downloads 3,292
Release Time : 2/24/2025

Model Overview

Through supervised fine-tuning (SFT) and model fusion techniques, this model excels in mathematical, programming, and scientific tasks, particularly outperforming some larger models in the field of mathematics.

Model Features

High-Performance Mathematical Reasoning
Outperforms 70B models on mathematical tasks (AIME 2024), approaching the performance of the full R1 model.
Multi-domain Fusion
Achieves comprehensive performance improvements by fusing specialized models in mathematics, programming, and scientific fields.
Open-Source Training Process
Provides a complete training and evaluation pipeline along with curated training datasets.

Model Capabilities

Mathematical reasoning
Programming task solutions
Scientific Q&A
Text generation

Use Cases

Education
Mathematical Problem Solving
Solves complex mathematical problems, such as integral calculations.
Achieved a score of 78.1 in the AIME 2024 test.
Programming
Programming Task Solutions
Answers programming-related questions and tasks.
Achieved a score of 61.6 in the LiveCodeBench test.
Scientific Research
Scientific Q&A
Answers questions in the scientific field.
Achieved a score of 65.0 in the GPQA-Diamond test.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase