# open-thoughts/open-thoughts

Fully open data curation for reasoning models

Repository: https://github.com/open-thoughts/open-thoughts
Canonical: https://ross.abutalabs.com/products/open-thoughts
Homepage: https://open-thoughts.ai
Language: Python
License: Apache-2.0
License Family: permissive
Topics: open-data, reasoning
Last push: 2025-12-02T21:57:07+00:00

## Health v2 (maintenance only)
Score: 45/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 55, release rhythm 35, longevity 41
- inputs: {"age_days": 583, "days_push": 274, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2323, forks 194 (observed 2026-08-28T04:06:37.283020+00:00)

## What it is
OpenThoughts is a community project curating fully open datasets for training reasoning models, including OpenThoughts3-1.2M and OpenThinker models that rival DeepSeek-R1 distills. It provides data curation recipes, training traces, and evaluation benchmarks for math and code reasoning.

## Use cases
- train a small reasoning model that beats DeepSeek-R1-Distill-Qwen-7B
- download open SFT reasoning traces for math and code
- fine-tune an LLM on curated chain-of-thought data
- reproduce a state-of-the-art reasoning dataset recipe
- evaluate reasoning models on open benchmarks
- build post-training data for agent models

## When to choose
- you need open, permissively licensed reasoning training data
- you want to distill reasoning capability into smaller models
- you are researching post-training data curation for LLMs

## When to avoid
- you need a ready-to-use inference server or chatbot product
- you want general-purpose pretraining corpora rather than reasoning SFT data
- you lack GPU resources for fine-tuning large models

## Facets
- artifact type: dataset
- maturity: active
- function: data-generation, machine-learning, llm-training, rag
- domain: large-language-models, machine-learning, artificial-intelligence, data-science
- platform: python
- tags: reasoning-datasets, post-training, sft, distillation, open-data, huggingface

## Member repositories
- open-thoughts/open-thoughts (main) score 45

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:37.283020+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:38:36.239407+00:00, confidence not recorded.
  - readme: https://github.com/open-thoughts/open-thoughts (fetched 2026-08-28T04:06:37.283020+00:00, sha d3dc2d558b93)
  - homepage: https://open-thoughts.ai (fetched 2026-08-29T10:18:50.546412+00:00, sha dbc73bf0ac18)
- Data as of 2026-08-30T08:39:29.467469+00:00.