Unverified50% confidenceFactExact time
Qwen2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
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Related Claims
UnverifiedThe Qwen 2.5 series was launched in 2024, covering a full range of open-source weights from 0.5B to 72B.75% similarUnverifiedQwen3于2025年4月发布，Dense模型引入QK RMS Norm，MoE模型取消共享专家并改用更细粒度专家切分配全局批量负载均衡损失，训练数据翻倍到36万亿Token72% similarUnverifiedQwen2 was released in mid-2024 and adopted architectural improvements such as Grouped Query Attention (GQA) and sliding window attention.72% similarUnverifiedQwen1.5于2024年2月发布，首次引入GQA和MoE版本，上下文窗口提升到32K69% similarUnverifiedQwen was officially opened to the public in 2023 and has continued to iterate.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/585434API
curl https://kongchang.com/api/v1/knowledge/claims/585434MCP
get_claim(id=585434)