DeepSeek-R1

DeepSeek Released January 2025

DeepSeek's reasoning model competing with OpenAI o1, released under MIT license

DeepSeek-R1

DeepSeekJanuary 2025

Training Data

Up to December 2024

DeepSeek-R1

January 2025

Parameters

671 billion (37B active)

Training Method

Advanced Reinforcement Learning

Context Window

128,000 tokens

Knowledge Cutoff

December 2024

Key Features

Open Source (MIT) • Advanced Reasoning • Cost Effective • Step-by-step Thinking

Capabilities

Reasoning: Outstanding

Math: Outstanding

Cost: Excellent

What's New in This Version

Competitive with o1 at fraction of the cost, open-sourced under MIT license

DeepSeek's reasoning model competing with OpenAI o1, released under MIT license

What's New in This Version

Competitive with o1 at fraction of the cost, open-sourced under MIT license

Technical Specifications

Parameters 671 billion (37B active)
Context Window 128,000 tokens
Training Method Advanced Reinforcement Learning
Knowledge Cutoff December 2024
Training Data Up to December 2024

Key Features

Open Source (MIT) Advanced Reasoning Cost Effective Step-by-step Thinking

Capabilities

Reasoning: Outstanding
Math: Outstanding
Cost: Excellent

Other DeepSeek Models

Explore more models from DeepSeek