# vLLM explains K3.1 attention and serving architecture

> Original title: K3.1 combines DSA sparse attention with recurrent Kimi Delta Attention.

- Source: vLLM
- Published: 2026-11-04T16:39:53.000Z
- Bad Influence: https://badinfluence.news/items/px9av7aemipuz8zffte0sh458
- Original: https://x.com/vllm_project/status/2118020419722834256

## Summary

vLLM says K3.1 combines DSA sparse attention with recurrent Kimi Delta Attention, storing MLA latents, indexer keys and KDA state in one packed cache. The serving work adds token-sharded prefill, exact top-k merging for decode context parallelism, and a fused low-latency decode path.

## Post

K3.1 combines DSA sparse attention with recurrent Kimi Delta Attention.

One packed cache holds MLA latents, indexer keys and KDA state. The serving work adds token-sharded prefill, exact top-k merging for decode context parallelism, and a fused low-latency decode path.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWS4WSea8AAYZJg.jpg&mode=thumb&exp=1794009600&sig=5d9b7da6695b6793)