deepseek-ai/deepseek-r1

★ 92,014Audience · developerComplexity · 3/5QuietLicenseSetup · moderate

Mindmap

mindmap
  root((repo))
    What it does
      Step-by-step reasoning
      Math and code solving
      Long chain of thought
    Model family
      DeepSeek-R1-Zero
      DeepSeek-R1 full
      Distilled versions
    Sizes available
      1.5B parameters
      Up to 70B parameters
      Local inference
    Training approach
      Reinforcement learning
      Cold-start data
      Self-checking answers
    Performance
      Math benchmarks
      Code benchmarks
      Reasoning tasks
    Getting started
      Download from Hugging Face
      Run locally
      Fine-tune smaller models

mindmap root((repo)) What it does Step-by-step reasoning Math and code solving Long chain of thought Model family DeepSeek-R1-Zero DeepSeek-R1 full Distilled versions Sizes available 1.5B parameters Up to 70B parameters Local inference Training approach Reinforcement learning Cold-start data Self-checking answers Performance Math benchmarks Code benchmarks Reasoning tasks Getting started Download from Hugging Face Run locally Fine-tune smaller models

Things people build with this

USE CASE 1

Download and run reasoning models locally on your own hardware without cloud API costs.

USE CASE 2

Fine-tune the smaller distilled models (1.5B, 70B) on your own math, coding, or reasoning tasks.

USE CASE 3

Study the paper and training methods to understand how reinforcement learning improves reasoning in language models.

USE CASE 4

Compare DeepSeek-R1 performance against other reasoning models on benchmark datasets.

Tech stack

Large Language ModelsReinforcement LearningHugging FaceModel Distillation

Getting it running

Difficulty · moderate Time to first run · 30min

Requires downloading large model weights from Hugging Face; GPU recommended for reasonable inference speed.

Use freely for any purpose, including commercial use, as long as you keep the copyright notice.

In plain English

DeepSeek-R1 is the public release of a family of large language models built by DeepSeek AI that are designed to be good at step-by-step reasoning, solving maths problems, writing code, and working through long chains of thought before producing an answer. The repository contains documentation, evaluation results, an accompanying paper, and links to download the actual model weights from Hugging Face. It does not contain the model itself as code; the heavy machine-learning weights live separately and are loaded by other software. The README describes the family in two parts. The first is the post-training method: rather than the usual approach of teaching the model with curated example answers (supervised fine-tuning) before reinforcement learning, the team applied reinforcement learning directly to a base model. The result, called DeepSeek-R1-Zero, learned to produce long chains of thought and self-check its answers, but suffered from issues like repetition and language mixing. DeepSeek-R1 adds "cold-start" data and additional stages to fix those issues. According to the README, DeepSeek-R1 reaches performance comparable to OpenAI-o1 on maths, code, and reasoning benchmarks. The second part is distillation: the team used data produced by DeepSeek-R1 to fine-tune smaller open-source models in sizes ranging from 1.5B up to 70B parameters, so users with less computing power can still benefit. Someone might use this repository to download the weights, run the models locally, study the paper, or fine-tune the smaller distilled checkpoints for specific tasks. The project is released under the MIT licence.

Copy-paste prompts

Prompt 1

How do I download and run DeepSeek-R1 locally? What hardware do I need for the different model sizes?

Prompt 2

Show me how to fine-tune one of the smaller DeepSeek-R1 distilled models on my own dataset using Hugging Face transformers.

Prompt 3

What is the difference between DeepSeek-R1-Zero and DeepSeek-R1? Why did they add cold-start data?

Prompt 4

How does DeepSeek-R1's reinforcement learning approach differ from supervised fine-tuning followed by RL?

Prompt 5

Which DeepSeek-R1 model size should I use if I want to run inference on a single GPU with 24GB VRAM?

Open on GitHub → Explain another repo

Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.