# EdgeRock-4B open-sources fixed-memory reasoning with 99.1% smaller KV Cache

> Original title: KV Cache砍掉99%，单卡吞吐飙6倍，港科CMU联手岩山科技开源新模型

- Source: QbitAI
- Published: 2026-10-26T06:45:25.000Z
- Bad Influence: https://badinfluence.news/items/u65me56ozaq6wfoh2v71smwhc
- Original: https://www.qbitai.com/2026/10/505407.html

## Summary

Rockmatrix AI, under Yanshan Technology, and collaborators including HKUST and CMU open-sourced EdgeRock-4B, which keeps KV Cache at 72 MiB even at 256K context. The article says a single 80GB GPU runs 512 concurrent long-reasoning sequences at 12,124 tokens/s, 6.42 times Qwen3.5-4B's best, while averaging 82.9% on AIME 2024–2026, according to the report.