DeepSeek-R1

weights

open(MIT, on HF

data

closed(RL/SFT training data not released)

code

partial(inference

license

MIT(OSI-style permissive, no use restrictions)

MIT-licensed open weights that explicitly permit distillation, but the RL and SFT post-training data and the full training code are not released, so this is open weights rather than open source.

• https://huggingface.co/deepseek-ai/DeepSeek-R1 recorded 2026-09-29

  The card states the model weights are licensed under the MIT License and that "DeepSeek-R1 series support commercial use, allow for any modifications and derivative works, including, but not limited to, distillation for training other LLMs"; the repo state shows "license:mit". Searched the body for training data, training pipeline and RL code releases: none, only evaluation tables and a model-download table.

• https://huggingface.co/deepseek-ai/DeepSeek-R1-0528 recorded 2026-09-29

  The R1-0528 card states "The use of DeepSeek-R1 models is also subject to" the MIT License and that the series "supports commercial use and distillation"; the repo state shows "license:mit".

Adoption is measured as Hugging Face downloads on the base R1 repository, summed with the declared R1-0528 SKU. That count leaves out the large family of R1 distills and the DeepSeek app and API surface, so it captures only part of the release's total use, which drove widespread open-reasoning adoption.

• https://huggingface.co/deepseek-ai/DeepSeek-R1 recorded 2026-08-13

  8,316,405 downloads in the trailing 30 days

DeepSeek-R1 was a frontier open reasoning model at its January 2025 release, matching or beating OpenAI's o1 on AIME, MATH-500 and LiveCodeBench; the figures are vendor-reported on its Hugging Face card. Newer DeepSeek releases, including the V4 family, have since surpassed it.

• https://huggingface.co/api/models?author=deepseek-ai&sort=createdAt&direction=-1&limit=25 recorded 2026-09-29

  Newest-first listing of the deepseek-ai organization lists "deepseek-ai/DeepSeek-V4-Pro" and "deepseek-ai/DeepSeek-V4-Flash" (created 2026-04-22) and "deepseek-ai/DeepSeek-V3.2"; searched the 25 newest entries for a DeepSeek-R2 repository: none.

• https://huggingface.co/deepseek-ai/DeepSeek-R1 recorded 2026-09-29

  The card benchmark table lists "AIME 2024 (Pass@1)", "MATH-500 (Pass@1)", "LiveCodeBench (Pass@1-COT)" and "SWE Verified (Resolved)" rows with the R1 column at 79.8, 97.3, 65.9 and 49.2, plus a GPQA Diamond value of 71.5 and MMLU 90.8, all vendor-reported against OpenAI o1.