Skip to content

Dolphin R1

A reasoning dataset with 800,000 samples

Visit official site
Category AI models

What Dolphin R1 is

The dataset is described as a Cognitive Computations release for training reasoning models, combining reasoning traces and chat samples into one resource.

What you can do with it

  • Train a model on complex reasoning
  • Draw on varied reasoning styles
  • Run large-scale training work

Getting started

  • Download the dataset from Hugging Face
  • Unpack files and inspect the structure
  • Split it into train, validation and test sets

Pros & cons

✓ What we like

  • Diverse reasoning styles and chat scenarios are included
  • Large-scale training is supported

! What to watch out for

  • The release is a dataset rather than a ready model
  • Sample counts come from the three named sources

FAQ

How large is the dataset?

The material states 800,000 samples in total.

Who created it?

The material names the Cognitive Computations team.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review