Honeyindex  /  LLMs  /  DeepSeek R1

DeepSeek R1

Breakthrough open-weights reasoning model

Context Window
128K tokens
Parameters
671B (MoE)
Source
Open
Modality
Text, Code

DeepSeek R1 was the breakthrough reasoning model demonstrating ChatGPT-level reasoning at significantly lower training costs. Open weights.

Specifications

Technical details
Developer
DeepSeek
Model Family
DeepSeek R1
Parameters
671B (MoE)
Context Window
128,000 tokens (128K)
Modality
Text, Code
Open Source
Yes
License
MIT
API Available
Yes
Release Date
January 20, 2025
Pricing
Free to self-host; $0.55/$2.19 via API