Search
- included,
1 result · page 1 of 1
- DeepSeek_Core
The character is modeled after the DeepSeek-V2 large language model (LLM), which utilizes a Mixture-of-Experts (MoE) architecture (236B total parameters, 21B active per token) and Multi-head Latent Attention (MLA). This core efficiency tran…
Chub990 tokens