Grade School Math 8K

GSM8K tests multi-step arithmetic in grade-school word problems. OpenAI introduced the dataset in 2021; a high score on this fixed test is one piece of evidence about mathematical reasoning.

6 models

Higher is better.

1Ternary Bonsai 2 27BPrism ML96.7
2Llama 3.1 70BMeta95.1
3Celeris-1Celeris93.7
4Phi-3.5-miniMicrosoft86.2
5Soofi S 30B-A3BSoofi Project86.1
6Llama 3.1 8BMeta84.5

Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric: Exact match
Tasks: Grade-school math word problems
Authors: Karl Cobbe and colleagues (OpenAI) (2021)
Source: Training Verifiers to Solve Math Word Problems ↗

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.