# DeepSeek open-sources DeepInfer, an inference engine for DeepSeek-V4.1-Flash

> Original title: DeepSeek 又一开源重器！深挖 DeepInfer 究竟藏着多少技术狠活？

- Source: Leiphone
- Published: 2026-10-29T14:14:00.000Z
- Bad Influence: https://badinfluence.news/items/o11p9p8pd0gt8kvje7l7czawy
- Original: https://www.leiphone.com/category/ai/HecXToP0EHbPjPd0.html

## Summary

DeepSeek has released DeepInfer on GitHub, the inference engine behind DeepSeek-V4.1-Flash, covering MoE expert parallelism, asynchronous batch scheduling, custom kernels, and speculative decoding. The source says a single cluster processes about 155 trillion tokens per day, and only offline inference is open, while online KV Cache storage and cross-replica load balancing are not yet fully released.

## Why it matters

The article walks through DeepInfer's asynchronous scheduling, state correction, and multi-GPU coordination, giving engineers a concrete view of how large-scale inference handles them.