Event ·Single source

DFM Mimir v1:10亿参数HRM开源模型仅用许可数据达前沿性能

DFM Mimir v1是丹麦基础模型团队发布的10亿参数语言模型,基于HRM架构从零训练。该模型仅使用许可的后训练数据,混合了161个数据集进行训练。在20个英语、数学与代码及丹麦语基准上,Mimir v1超过HRM-Text 1B,并可与Qwen 3.5 4B和Gemma 4 E2B竞争。它在丹麦语任务上刷新了最先进水平,模型现已在Hugging Face Hub开放下载。

Current read

丹麦团队开源了10亿参数的Mimir v1,只用合规数据就追上4B模型,丹麦语还是第一。

2 sources55° AI heatLast updated 2026/8/13 17:37:53

Evidence

2 sources

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.