O

Openr1 Distill 7B

Developed by open-r1
OpenR1-Distill-7B is a post-trained version of Qwen2.5-Math-7B on the Mixture-of-Thoughts dataset, designed to teach language models step-by-step reasoning.
Downloads 134
Release Time : 5/22/2025

Model Overview

This model replicates the reasoning capabilities of DeepSeek-R1-Distill-Qwen-7B while maintaining full openness and reproducibility, making it suitable for research on reinforcement learning with verifiable rewards (RLVR) during reasoning.

Model Features

Step-by-step Reasoning
Trained on the Mixture-of-Thoughts dataset, the model is capable of complex step-by-step reasoning.
Open Reproducibility
Fully open datasets and training methods ensure reproducible results.
Long Context Support
RoPE base frequency extended to 300k supports training with 32k context length.

Model Capabilities

Mathematical problem-solving
Programming task resolution
Scientific problem reasoning
Multi-step reasoning generation
Long-text comprehension

Use Cases

Education
Mathematical Problem Solving
Helps students understand and solve complex mathematical problems.
Achieved 89.0% accuracy on the MATH-500 benchmark.
Research
Reasoning-time Computation Research
Used for studying reinforcement learning with verifiable rewards (RLVR) during reasoning.
Programming
Code Generation and Understanding
Assists developers in generating and understanding complex code.
Achieved 39.4% accuracy on LiveCodeBench v5.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase