Unverified50% confidenceFactExact time
nomic-embed-text嵌入模型有137M参数，all-MiniLM-L6-v2有22M参数，BGE-small为24M-109M参数
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
UnverifiedMiniMax M3 每个 token 推理时约激活 23B 参数67% similarUnverified训练7B参数模型的全参数微调通常需要至少4张A100（80G）显卡，而LoRA使单张RTX 4090（24G）完成7B量级模型微调成为可能65% similarUnverifiedLLaMA 2的32000词表在4096维隐层下嵌入层需占用约1.3亿参数65% similarUnverifiedLFM2.5-8B-A1B总参数量为8B，推理时仅激活1.5B参数65% similarVerifiedWhisper提供从tiny（39M参数）到large-v3（1.5B参数）的多个规格65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/533255API
curl https://kongchang.com/api/v1/knowledge/claims/533255MCP
get_claim(id=533255)