xtools-at/precious-datasets/Claude-Distills - Storage Bucket

Claude-Distills

A curated collection of open-source Claude distillation datasets, unified and deduplicated.

Note: This repo only provides unified formatting, deduplication, and documentation. All credits go to the original data creators. I did NOT create any of the original data.

Data Sources

Source	Samples	Description
claude-sonnet-4.6-120000x	119,446	Claude Sonnet 4.6 general, code, math, psychology data
claude-opus-4.6-10000x	9,633	Claude Opus 4.6 multi-task data
claude-opus-4.6-high-reasoning-700x	758	High-quality reasoning data
Claude-Opus-4.6-Reasoning-887x	887	Reasoning task data
Claude-Opus-4.6-Reasoning-500x	500	Reasoning task data
Claude-Sonnet-X-Opus-4.6-Reasoning-small-500	524	Sonnet & Opus mixed data
claude-4.5-opus-high-reasoning-250x	250	Claude 4.5 Opus reasoning data
claude-opus-4.6-4.7-reasoning-8.7k	8,706	Claude Opus 4.6/4.7 multi-category reasoning data
Total (after dedup)	140,504

What This Repo Does

1. Unified formatting: Converted all data sources to a consistent messages format with source field
2. Deduplication: Removed 3,098 duplicate samples by user prompt content
3. Documentation: Added data statistics, source attribution, and usage instructions

Data Format

{
  "messages": [
    {"role": "system", "content": "You are a helpful AI assistant."},
    {"role": "user", "content": "Question content"},
    {"role": "assistant", "content": "thinking process and answer"}
  ],
  "source": "claude-sonnet-4.6-120000x"
}

Thinking Process Format

Assistant responses include thinking process using special tokens to mark the thinking section, followed by the final answer.

Statistics

Metric	Value
Total samples	140,504
From Sonnet 4.6	119,446 (85.0%)
From Opus datasets	12,354 (8.8%)
From Opus 4.6/4.7 Reasoning	8,704 (6.2%)
Duplicates removed	3,100

Usage

from datasets import load_dataset

dataset = load_dataset("Kassadin88/Claude-Distillation-Dataset")

License

This dataset is for research and educational purposes only. Please follow the terms of use of the original data sources.

Citation

If you use this dataset, please cite:

@misc{claude-distills,
  title={Claude-Distills: A Curated Collection of Open-Source Claude Distillation Datasets},
  author={Kassadin88},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/Kassadin88/Claude-Distillation-Dataset}
}

And the original sources:

@misc{claude-sonnet-4.6-120000x,
  title={claude-sonnet-4.6-120000x},
  author={Roman1111111},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/Roman1111111/claude-sonnet-4.6-120000x}
}

@misc{claude-opus-4.6-10000x,
  title={claude-opus-4.6-10000x},
  author={Roman1111111},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/Roman1111111/claude-opus-4.6-10000x}
}

@misc{claude-opus-4.6-high-reasoning-700x,
  title={claude-opus-4.6-high-reasoning-700x},
  author={dalisoft},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/dalisoft/claude-opus-4.6-high-reasoning-700x}
}

@misc{claude-opus-4.6-reasoning-887x,
  title={Claude-Opus-4.6-Reasoning-887x},
  author={TeichAI},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/TeichAI/Claude-Opus-4.6-Reasoning-887x}
}

@misc{claude-sonnet-x-opus-4.6-reasoning-small-500,
  title={Claude-Sonnet-X-Opus-4.6-Reasoning-small-500},
  author={Hastagaras},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/Hastagaras/Claude-Sonnet-X-Opus-4.6-Reasoning-small-500}
}

@misc{claude-4.5-opus-high-reasoning-250x,
  title={claude-4.5-opus-high-reasoning-250x},
  author={TeichAI},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/TeichAI/claude-4.5-opus-high-reasoning-250x}
}

@misc{claude-opus-4.6-4.7-reasoning-8.7k,
  title={claude-opus-4.6-4.7-reasoning-8.7k},
  author={angrygiraffe},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k}
}

Acknowledgments

Thanks to all original data creators and providers. This is just a curated collection of their work.