TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

DeepSeek-V3-0324 released, 641GB, MIT licensed, >20tok/SEC on $<10k hardware

15 点作者 Thoreandan大约 2 个月前

4 条评论

42772827大约 2 个月前
The headline seems to imply that the full 641GB model is running at &gt;20tok&#x2F;sec on the Mac Studio, but the blog says:<p>&gt;The model only came out a few hours ago and MLX developer Awni Hannun already has it running at &gt;20 tokens&#x2F;second on a 512GB M3 Ultra Mac Studio ($9,499 of ostensibly consumer-grade hardware) via mlx-lm and this mlx-community&#x2F;DeepSeek-V3-0324-4bit 4bit quantization, which reduces the on-disk size to 352 GB.
mdaniel大约 2 个月前
the non-archive.org submission, where simonw actually is watching and commenting <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=43462317">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=43462317</a>
sinenomine大约 2 个月前
Impressive use of reasoner CoT distillation method applied to deepseek R1. MIT license for the weights. Thanks, Deepseek!
ilrwbwrkhv大约 2 个月前
The only open AI in town