Bluesky Thread

Day 1 of DeepSeek open source week: FlashMLA

February 24, 2025 View original thread

Day 1 of DeepSeek open source week: FlashMLA

MLA=Multihead Latent Attention

One of the big innovations that made V3 such a notable model

github.com/deepseek-ai/...

github.com

GitHub - deepseek-ai/FlashMLA

Contribute to deepseek-ai/FlashMLA development by creating an account on GitHub.

21 1

fyi here’s a good overview of MLA

medium.com/towards-data...

medium.com

DeepSeek-V3 Explained 1: Multi-head Latent Attention

Key architecture innovation behind DeepSeek-V2 and DeepSeek-V3 for faster inference

it’s MIT license.. if deepseek flexes any more they’re gonna tear a muscle

More like this