DeepSeek-V2
DeepSeek's flagship MoE model with exceptional efficiency
DeepSeek-V2
DeepSeek • May 2024
Training Data
Up to early 2024
DeepSeek-V2
May 2024
Parameters
236 billion (21B active)
Training Method
Mixture of Experts
Context Window
128,000 tokens
Knowledge Cutoff
February 2024
Key Features
Mixture of Experts • High Efficiency • Open Source
Capabilities
Reasoning: Excellent
Coding: Very Good
Math: Excellent
What's New in This Version
Exceptional performance with MoE architecture efficiency
DeepSeek's flagship MoE model with exceptional efficiency
What's New in This Version
Exceptional performance with MoE architecture efficiency
Technical Specifications
Key Features
Capabilities
Other DeepSeek Models
Explore more models from DeepSeek
DeepSeek-V4-Pro
DeepSeek's frontier MoE flagship closing the gap with leading proprietary models on reasoning and agentic coding
DeepSeek-V4-Flash
DeepSeek's smaller, fast variant of V4 — same architecture at a fraction of the cost and latency
DeepSeek-V3.2
DeepSeek's latest flagship model matching GPT-5 performance with integrated tool-use thinking
