Unverified60% confidenceFactTime unknown
DeepSeek 8B adopts the GQA (Grouped Query Attention) attention mechanism to reduce VRAM usage during inference.
2
Sources
60%
Confidence
Long-term
Relevance
6/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe model introduces QSA, a self-developed absorptive Attention mechanism from Tongyi Qianwen, which is highly similar in concept to DeepSeek's DSA (DeepSeek Sparse Attention).78% similarUnverifiedDeepSeek's innovations in Multi-head Latent Attention (MLA) and FP8 mixed-precision training compress KV Cache memory usage to 5%–13% of standard multi-head attention.74% similarVerifiedDeepSeek V4 adopts a proprietary hierarchical compressed attention strategy, featuring two alternating attention modes: CSA (Compressed Sink Attention) and HCA (Heavy Compressed Attention).72% similarUnverifiedDeepSeek-V3.2-Exp首次引入自研的DeepSeek Sparse Attention（DSA）稀疏注意力技术72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/36397API
curl https://kongchang.com/api/v1/knowledge/claims/36397MCP
get_claim(id=36397)