open-r1

Overview

Open-source reproduction of DeepSeek-R1's reasoning architecture, implemented in Python. Enables builders to run and fine-tune chain-of-thought reasoning models locally without proprietary constraints. Designed for transparency and reproducibility of the R1 approach.

Best for

Best for
Researchers and builders needing transparent, locally-controlled reasoning models

Use cases

• Fine-tuning reasoning models on custom datasets
• Running inference locally for reasoning-heavy tasks
• Studying and modifying chain-of-thought mechanisms

Notes

Open-source reproduction of DeepSeek-R1’s reasoning architecture, implemented in Python. Enables builders to run and fine-tune chain-of-thought reasoning models locally without proprietary constraints. Designed for transparency and reproducibility of the R1 approach.

26,029 stars on GitHub. Last updated 2026-04-02. Licensed Apache-2.0.

Use cases

• Fine-tuning reasoning models on custom datasets
• Running inference locally for reasoning-heavy tasks
• Studying and modifying chain-of-thought mechanisms

Pros

• Fully open source with no licensing restrictions
• Community-maintained with 26k+ GitHub stars indicating active development
• Enables local deployment without external API dependencies

Cons

• Requires significant computational resources to run effectively
• Community project, not backed by commercial support
• May lag behind proprietary R1 improvements and optimizations

Indexed from awesome-llm and enriched against its public facts.