DeepSeek-V3-0324
DeepSeek's updated V3 model with improved post-training using R1 techniques
DeepSeek-V3-0324
DeepSeek • March 2025
Training Data
Up to March 2025
DeepSeek-V3-0324
March 2025
Parameters
671 billion (37B active)
Training Method
Advanced MoE with R1 RL techniques
Context Window
128,000 tokens
Knowledge Cutoff
March 2025
Key Features
Improved Post-training • R1-enhanced • Advanced Performance
Capabilities
Reasoning: Outstanding
Coding: Excellent
Efficiency: Outstanding
What's New in This Version
Enhanced post-training pipeline incorporating lessons from DeepSeek-R1
DeepSeek's updated V3 model with improved post-training using R1 techniques
What's New in This Version
Enhanced post-training pipeline incorporating lessons from DeepSeek-R1
Technical Specifications
Key Features
Capabilities
Other DeepSeek Models
Explore more models from DeepSeek
DeepSeek-V2
DeepSeek's flagship MoE model with exceptional efficiency
DeepSeek Coder V2
DeepSeek's specialized coding model with advanced programming capabilities
DeepSeek-V3
DeepSeek's latest flagship model with enhanced capabilities and efficiency