I swear DeepSeek open-sourcing everything is some Sun-Tzu shit. America is trying to build trillio...
TL;DR · AI Summary
DeepSeek 全面开源 V4 系列模型（含 1M 上下文、1.6T 参数 Pro 版），被解读为以‘孙子兵法’式战略瓦解美国 AI 垄断逻辑：当 AGI 核心技术公开，闭源护城河崩塌，估值与商业模型将重构。
Key Takeaways
- DeepSeek-V4-Pro 以 1.6T 总参数、49B 活跃参数实现媲美顶级闭源模型的性能，并完全开源。
- AGI ‘秘方’公开化将削弱闭源企业的护城河，使高估值难以为继，商业模式被迫转向算力部署效率。
- 中美AI竞争已从技术竞赛升级为范式之争：垄断封闭 vs 开源协同，胜负关键或将取决于 token/s 的实际交付能力。
Outline
Jump quickly between sections.
Mindmap
See how the topics connect at a glance.
查看大纲文本（无障碍 / 无 JS 友好）
- DeepSeek 开源战略
- 技术层
- V4-Pro：1.6T/49B，1M context
- V4-Flash：284B/13B，轻量高效
- 战略层
- Sun-Tzu 式非对称破壁
- 瓦解闭源估值逻辑
- 产业层
- 算力即新护城河
- token/s 成核心KPI
Highlights
Key sentences worth saving and sharing.
I swear DeepSeek open-sourcing everything is some Sun-Tzu shit.
If the secret recipe to AGI is (eventually) public, the business model changes. Valuations compress.
We need to put all of our resources into securing compute, because in the end, this will come down to who can deploy the most tokens per second.
America is trying to build trillion-dollar AI monopolies, and China is trying to make that impossible.
If the secret recipe to AGI is (eventually) public, the business model changes. Valuations compress." / X
AI Breakfast on X: "I swear DeepSeek open-sourcing everything is some Sun-Tzu shit. America is trying to build trillion-dollar AI monopolies, and China is trying to make that impossible. If the secret recipe to AGI is (eventually) public, the business model changes. Valuations compress." / X
Don’t miss what’s happening
I swear DeepSeek open-sourcing everything is some Sun-Tzu shit. America is trying to build trillion-dollar AI monopolies, and China is trying to make that impossible. If the secret recipe to AGI is (eventually) public, the business model changes. Valuations compress. Defensibility disappears. The premium on being closed starts to look like a liability instead of a strength. We need to put all of our resources into securing compute, because in the end, this will come down to who can deploy the most tokens per second.
Quote
DeepSeek
@deepseek_ai
·
Apr 24
DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. 
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. DeepSeek-V4-Flash: 284B total / 13B active params.
·
51
157
1.1K
175
Read 51 replies