# SGLang announces day-zero support for Kimi K3.1 Pro with sparse attention

> Original title: Kimi K3.1 Pro replaces dense MLA with sparse attention (DSA). It's much more efficient and easier to serve.

- Source: SGLang
- Published: 2026-11-04T16:26:53.000Z
- Bad Influence: https://badinfluence.news/items/q18gejmglzlz4k4kaomarw5v4
- Original: https://x.com/sgl_project/status/2118017145548451860

## Summary

SGLang says it supports Kimi K3.1 Pro on day zero, after the model replaces dense MLA with sparse attention (DSA). It reports up to 33.7K tokens per second per GPU on AgentX with GB300 and 450 tokens per second for a single user with DFlash2 and MegaKernel. Support covers NVIDIA Vera Rubin, Blackwell and Hopper, AMD, TPU, Trainium and PPU.

## Why it matters

The post names the architecture change and the specific throughput and hardware support, so readers can judge how serving the model differs from the dense MLA version.

## Post

Kimi K3.1 Pro replaces dense MLA with sparse attention (DSA). It's much more efficient and easier to serve.

SGLang delivers on Day 0:
· Up to 33.7K tok/s per GPU (AgentX, GB300)
· 450 tok/s single user with DFlash2 + MegaKernel
· NVIDIA Vera Rubin / Blackwell / Hopper, AMD, TPU, Trainium, PPU

[Video](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Famplify_video_thumb%2F2118015453582995458%2Fimg%2Fh3dUZOvtEJYazdbj.jpg&mode=thumb&exp=1794009600&sig=9e6c5fa47bb17281)

## Quoted post @Kimi_Moonshot

> Releasing the model weights of Kimi K3.1 Pro: https://huggingface.co/moonshotai/Kimi-K3.1-Pro
> We'll keep advancing open intelligence.

https://x.com/Kimi_Moonshot/status/2118010528123773297