Skip to content

Dataset · open-thoughts

Synthetic Reasoning Data: 114k Examples

open-thoughts/OpenThoughts-114k

114k reasoning examples spanning math, science, code, and puzzles for training LLMs.

What's inside

OpenThoughts-114k is an open synthetic reasoning dataset with 114k examples. It covers mathematics, science, programming, and puzzles, and was used to train OpenThinker-7B and OpenThinker-32B models. A ready-to-train default subset is available; the data is text-based.

What you can do with it

  • SFT LLMs for reasoning tasks
  • Fine-tune LoRA on math and code
  • Evaluate models on puzzles and science
Size
100K–1M rows
Downloads this month
144.3K
Likes
945

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("open-thoughts/OpenThoughts-114k")