GSM8K Math Reasoning Dataset

GSM8K Mathematical Reasoning
Dataset

GSM8K (Grade School Math 8K) is a high-quality benchmark dataset of elementary school math word problems created by OpenAI. The main configuration contains 8,792 problems with detailed step-by-step solutions and has become the standard benchmark for evaluating the mathematical reasoning capabilities of large language models. Each problem requires 2-8 steps of basic mathematical operations to solve.

Get the Dataset Now
8,792 Problems Step-by-Step Solutions MIT License OpenAI Benchmark
🧮
8,792
Total Math Problems
📝
2-8 Steps
Solution Step Range
🏛️
OpenAI
Dataset Creator
📜
MIT
Open License

Dataset Highlights

The gold-standard benchmark dataset for evaluating mathematical reasoning abilities of large language models

🎓

Elementary Math Problems

Problems cover basic elementary-level mathematical operations, including addition, subtraction, multiplication, division, fractions, percentages, and more. No advanced mathematics knowledge is required, focusing on testing reasoning rather than computational complexity.

📐

Step-by-Step Solutions

Each problem includes a detailed step-by-step solution process, with intermediate calculation steps and final answers annotated, providing high-quality labeled data for Chain-of-Thought training.

🌍

Language Diversity

Problems use natural language to describe a rich variety of real-life scenarios, covering topics such as shopping, speed, and age. Flexible language avoids formulaic patterns and better reflects real-world problems.

🔗

Multi-Step Reasoning

Each problem requires 2-8 reasoning steps to solve, demanding coherent logical reasoning and intermediate-state tracking capabilities from models, effectively distinguishing the reasoning levels of different models.

💬

Natural Language Problems

All problems are presented in natural language text, without formulas or symbolic expressions, testing a model's ability to understand mathematical relationships from text and extract key numerical values.

📊

Standard Benchmark

Widely used by mainstream large language models such as GPT-4, Claude, and Gemini to evaluate mathematical reasoning abilities, it is one of the most frequently cited mathematics benchmarks in academic papers and technical reports.

Use Cases

Comprehensively support mathematical reasoning research, from model evaluation to educational R&D

🧠

Mathematical Reasoning Evaluation

Evaluate the multi-step mathematical reasoning capabilities of large language models, quantifying accuracy and reasoning quality on elementary math problems

⛓️

Chain-of-Thought Training

Use step-by-step solution data to train Chain-of-Thought reasoning, improving models' ability to break down complex problems step by step

📈

LLM Benchmarking

Compare the mathematical reasoning performance of different models using a standardized benchmark, and track capability trends across model iterations

🏫

Educational AI Development

Build intelligent math tutoring systems and automated problem-solving tools, providing training and evaluation data for AI applications in K-12 education

math reasoning benchmark education OpenAI

Real Data Preview

View limited samples verified against the delivery files in the “Dataset” tab on this page. Delivery includes ZIP files for the GSM8K main configuration training and test sets, along with the original license and source manifest.

View Samples and Fields

Get Started in 3 Quick Steps

From browsing to use, start your mathematical reasoning research in just minutes

01

Browse the Dataset

View GSM8K dataset details on the Ace Data Cloud platform to learn about metadata such as problem formats, solution structures, and license terms.

02

Get the Data

Choose a one-time package, then download ZIP files for the main configuration training and test sets from “My Datasets” after payment.

03

Evaluate and Train

Use the dataset to evaluate model mathematical reasoning capabilities, or as a training data source for Chain-of-Thought fine-tuning.

Get the Dataset

Start Exploring GSM8K Mathematical Reasoning Data

A standard mathematical reasoning benchmark by OpenAI, licensed under MIT and delivered upon request according to the displayed plans. Whether you are evaluating large model capabilities or training reasoning models, GSM8K is an indispensable dataset.

Get the Dataset Now