1
智谱开源 GLM-5.3 模型权重，主打智能体编程与网络防御
IT之家 8 月 29 日消息，智谱官方昨晚宣布 开源 GLM-5.3 模型权重 ，已正式开放下载，支持本地运行与个性化定制。该模型擅长复杂编码、防御性网络安全以及长程任务。 智谱 Z.ai 负责人李子玄表示，该模型使用需要遵循 GLM-5.3 许可协议，支持本地部署、模型微调及 商业化使用 。仅当…
2
阿里开放 Qwen3.8-2.4T-A95B 模型权重：2.4T MoE、激活 95B、原生 256K 上下文
千问首次开源超大MoE模型权重，2.4T参数激活95B，原生256K上下文，大模型玩家不容错过。
IT之家 8 月 13 日消息，阿里云“魔搭 ModelScope 社区”公众号昨天（12 日）深夜宣布，阿里 Qwen 团队正式开放 Qwen3.8-2.4T-A95B 模型权重。 此次是 Qwen-Max 级别的模型首次开源权重：模型总参数为 2.4T，每个 Token 激活 95B 参数， 原…
3
Show HN: Weightlift – the missing package manager for downloading model weights
模型权重下载管理利器，让Web AI应用轻松搞定模型下载、缓存与进度展示，告别重复造轮子。
If you’re building an on-device web-based AI app, you’ve probably built this 50 times already: Something that tells your users which models are availa…
4
No Time Like the Present: Agentic Test-Time Training for LLM Agents
提出Agentic Test-Time Training方法，让LLM智能体在长时任务中自适应调整权重，有效解决轨迹退化、策略失效难题。
arXiv:2607.03441v1 Announce Type: cross Abstract: LLM agents often degrade over long episodes: as trajectories grow, they revisit explored states, rep…
5
Ask HN: Smallest amount of working ML weights that can be tattooed on a body?
把机器学习模型权重纹身上？HN老哥算出了最小尺寸，硬核极客的浪漫。
Recently saw this comment on another HN thread about the US government gating access to GPT-5.6 and how it harkens back to the 1990s encryption-as-exp…
6
LatentSkill: From In-Context Textual Skills to In-Weight Latent Skills for LLM Agents
将Agent的上下文技能固化进模型权重，解锁大模型泛化新范式
arXiv:2606.06087v1 Announce Type: cross Abstract: Agent systems increasingly use textual skills to encode reusable task procedures, but injecting thes…
7
AI Agent that at inference time updates it's harness and model weights
在推理时自动更新模型权重和框架，SIA在ML竞赛中横扫所有对手夺得第一
Article URL: https://github.com/hexo-ai/sia Comments URL: https://news.ycombinator.com/item?id=48344807 Points: 4 # Comments: 0