Unverified50% confidenceFactExact time
Llama 3于2024年发布，参数规格为8B、70B和405B，训练数据量超过15万亿tokens，采用分组查询注意力（GQA）技术
1
Sources
50%
Confidence
Long-term
Relevance
8/12/2026
First Seen
Sources
Related Claims
UnverifiedMeta's Llama series has been open-sourced since 2023, with the Llama 3.1 version reaching a maximum parameter count of 405B.68% similarVerifiedLLaMA was released by Meta in 2023 and is a landmark project in open-source large language models, demonstrating that smaller models trained on high-quality data can match closed-source models with far more parameters.67% similarUnverifiedLlama 2于2023年7月发布，将上下文长度扩展至4096 tokens，并提供经过RLHF微调的Chat版本63% similarUnverifiedLlama 1于2023年2月发布，参数规模为7B至65B，采用纯Transformer解码器架构，引入RoPE位置编码和SwiGLU激活函数62% similarUnverifiedQwen2于2024年6月发布，将GQA铺到全系列，KV缓存节省40%，训练数据从3万亿Token增长到7万亿61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/733650API
curl https://kongchang.com/api/v1/knowledge/claims/733650MCP
get_claim(id=733650)