# Ryan Hanrui Wang on how KV-cache management shapes agent latency and cost

> Original title: Ryan Hanrui Wang (@nebiusai) on moving from fast tokens to fast agents.

- Source: LMSYS Org
- Published: 2026-11-09T23:15:41.000Z
- Bad Influence: https://badinfluence.news/items/vtirptreidwvd369rob16ikx0
- Original: https://x.com/lmsysorg/status/2119931964094574851

## Summary

Nebius AI's Ryan Hanrui Wang argues that moving from fast tokens to fast agents depends on KV-cache management, scheduling, and model selection. He says these choices determine context reuse, latency, and the economics of production agents.

## Post

Ryan Hanrui Wang (@nebiusai) on moving from fast tokens to fast agents.

How KV-cache management, scheduling and model selection shape context reuse, latency and the economics of production agents.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWuCFYNbUAAEpHH.jpg&mode=thumb&exp=1794528000&sig=3544ddfe6c7d6796)