LogoDeepSeek Model1 Insight
  • 首页
  • 什么是 MODEL1
  • 模型对比
  • 博客
  • 关于本站

社区动态

来自 Twitter 的 DeepSeek MODEL1 最新讨论和社区观点

LogoDeepSeek Model1 Insight

本站为独立信息与分析站点，与 DeepSeek 官方不存在任何隶属或合作关系。

产品
  • 功能
  • 价格
  • 常见问题
资源
  • 更新日志
  • 路线图
公司
  • 关于我们
  • 联系我们
  • 邮件列表
法律
  • Cookie政策
  • 隐私政策
  • 服务条款
© 2026 DeepSeek Model1 Insight. 版权所有最后更新: 2026-03-08
ZhuoSSS
TokenPark
@ZhuoSSS
Detail
A new DeepSeek model, Model 1, has quietly appeared on GitHub. ->On the first anniversary of the release of DeepSeek-R1, a mysterious Model1 model appeared in the FlashMLA codebase update, possibly the codename for a new model to be released soon; ->Code analysis shows that
A new DeepSeek model, Model 1, has quietly appeared on GitHub. ->On the first anniversary of the release of DeepSeek-R1, a mysterious Model1 model appeared in the FlashMLA codebase update, possibly the codename for a new model to be released soon; ->Code analysis shows that https://t.co/b78aIn5mVrA new DeepSeek model, Model 1, has quietly appeared on GitHub. ->On the first anniversary of the release of DeepSeek-R1, a mysterious Model1 model appeared in the FlashMLA codebase update, possibly the codename for a new model to be released soon; ->Code analysis shows that https://t.co/b78aIn5mVr
techtechchina
Tech Tech China
@techtechchina
Detail
🚨 #DeepSeek’s next model just leaked. GitHub code reveals “MODEL1” — a new architecture (not V3.2), with changes to KV cache layout, sparsity handling, and FP8 decoding. Launch rumored around Feb. Another DeepSeek moment incoming.
🚨 #DeepSeek’s next model just leaked. GitHub code reveals “MODEL1” — a new architecture (not V3.2), with changes to KV cache layout, sparsity handling, and FP8 decoding. Launch rumored around Feb. Another DeepSeek moment incoming. https://t.co/ItW6ACb6wG🚨 #DeepSeek’s next model just leaked. GitHub code reveals “MODEL1” — a new architecture (not V3.2), with changes to KV cache layout, sparsity handling, and FP8 decoding. Launch rumored around Feb. Another DeepSeek moment incoming. https://t.co/ItW6ACb6wG🚨 #DeepSeek’s next model just leaked. GitHub code reveals “MODEL1” — a new architecture (not V3.2), with changes to KV cache layout, sparsity handling, and FP8 decoding. Launch rumored around Feb. Another DeepSeek moment incoming. https://t.co/ItW6ACb6wG
imangegatehouse
gatehouse
@imangegatehouse
Detail
A model for contextually managing information states, Manifold-Constrained Hyper-Connections (mHC)/Engram—DeepSeek [v4?] stepping away from a linear versioning path? Structural re-architecture, not feature-layer refinement—me think. tao-hpu.medium.com/deepseeks-mode…
sumjitg
Sumjit
@sumjitg
Detail
DeepSeek might be dropping MODEL1 on the one-year anniversary of their R1 release someone found a GitHub code snippet referencing “MODEL1” in their flash attention library with new KV cache optimizations R1 was the Chinese open-source reasoning model that rivaled OpenAI’s o1
DeepSeek might be dropping MODEL1 on the one-year anniversary of their R1 release someone found a GitHub code snippet referencing “MODEL1” in their flash attention library with new KV cache optimizations R1 was the Chinese open-source reasoning model that rivaled OpenAI’s o1 https://t.co/oOit3ejESBDeepSeek might be dropping MODEL1 on the one-year anniversary of their R1 release someone found a GitHub code snippet referencing “MODEL1” in their flash attention library with new KV cache optimizations R1 was the Chinese open-source reasoning model that rivaled OpenAI’s o1 https://t.co/oOit3ejESB
PIPOnew_
PIPO.new
@PIPOnew_
Detail
DeepSeek Model 1 Emerges: One Year After R1 Shook the Industry On the anniversary of DeepSeek-R1's release that kicked off the open-source LLM revolution, mysterious references to 'Model1' appeared in DeepSeek's FlashMLA codebase. The community speculates this could be the
DeepSeek Model 1 Emerges: One Year After R1 Shook the Industry On the anniversary of DeepSeek-R1's release that kicked off the open-source LLM revolution, mysterious references to 'Model1' appeared in DeepSeek's FlashMLA codebase. The community speculates this could be the https://t.co/qmDUw4u1yV
VICOINDAO
拓哥 互关中
@VICOINDAO
Detail
DeepSeek新模型呼之欲出 “MODEL1”作为临时代号曝光 在DeepSeek-R1 发布一周年之际 DeepSeek 在GitHub更新FlashMLA内核代码，其中114个文件中有28处提到：MODEL1，MODEL是模型的意思。 这应该不是V3系列补丁，而是全新架构，很可能就是大家等了很久的V4（或内部代号）。
DeepSeek新模型呼之欲出 “MODEL1”作为临时代号曝光 在DeepSeek-R1 发布一周年之际 DeepSeek 在GitHub更新FlashMLA内核代码，其中114个文件中有28处提到：MODEL1，MODEL是模型的意思。 这应该不是V3系列补丁，而是全新架构，很可能就是大家等了很久的V4（或内部代号）。 https://t.co/lASG0b8aAS
chetaslua
Chetaslua
@chetaslua
Detail
Deepseek Model-1 is likely a new architecture > The model is even compatible with Nvidia's B200. DeepSeek is really fast! > DeepSeek updated its FlashMLA code on GitHub, mentioning MODEL1 in 28 out of 114 files, presenting it as a different model from V32.
Deepseek Model-1 is likely a new architecture > The model is even compatible with Nvidia's B200. DeepSeek is really fast! > DeepSeek updated its FlashMLA code on GitHub, mentioning MODEL1 in 28 out of 114 files, presenting it as a different model from V32. https://t.co/ktdKXLBcSU