DeepSeek V4 Pro: 1.6T Model Runs on Huawei, But Who's Buying?
Another day, another gigantic AI model lands with a thud. This time it's DeepSeek, rolling out their V4 Pro and V4 Flash models, and they've got a new trick up their sleeve: ditching NVIDIA for Huawei's Ascend chips. Color me intrigued, and deeply skeptical.
⚡ Key Takeaways
- DeepSeek has released V4 Pro (1.6T params) and V4 Flash (284B params) models with a 1M token context window. 𝕏
- The models are designed to run on Huawei Ascend chips, signaling a move away from NVIDIA hardware due to geopolitical factors. 𝕏
- Technical advancements in Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) are credited for improved efficiency and reduced memory usage. 𝕏
Worth sharing?
Get the best AI stories of the week in your inbox — no noise, no spam.
Originally reported by Latent Space
