0

GSM8K

Frontier

GSM8K is a classic math word problems dataset.

Domain
rl-env
License
unknown
Published
Jan 2026

Cite

Notes

Only stored in your browser.

Attribution

Leaderboard scores
OpenReward
Attribution policy →

Top score 96.4 by Claude 3.5 Sonnet - 37 models reporting (4 frontier)

Score history

4
0255075100Nov 22Apr 23Sep 23Feb 24Jul 24GPT-3.5 TurboGPT-4Claude 3.5 Sonnet

Top models

37
GSM8KBar chart with 21 bars. Highest value: Llama 3 405B at 96.8.
21 models

FAQ

What is GSM8K?
GSM8K is a classic math word problems dataset.
What is the current top score on GSM8K?
The top reported score is 96.4 by Claude 3.5 Sonnet, across 37 models reporting (4 from frontier labs).
What license is GSM8K under?
GSM8K is available under unknown.