English(EN) Rock Matrix Open-Sources EdgeRock-4B, Cutting Qwen3.5-4B's 256K KV Cache From 8 GiB to 72 MiB

Rock Matrix 开源 EdgeRock-4B，大幅减少 KV 缓存

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-27 02:52

Rock Matrix 已开源 EdgeRock-4B，该模型源自 Qwen3.5-4B。此新模型在保持 256K 上下文长度的同时，将 KV 缓存大小从 8 GiB 大幅削减至 72 MiB。EdgeRock-4B 的吞吐量是其前身的六倍多，对推理能力的影响极小。 AI

影响此次发布为边缘部署提供了更高效的模型，有望降低推理成本，并使其在资源受限的环境中得到更广泛的应用。

排序理由发布了具有性能改进的新开源模型。[lever_c_demoted from research: ic=1 ai=1.0]

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 45 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 发布了具有性能改进的新开源模型。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: model release, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

Pandaily TIER_1 English(EN) · [email protected] (Pandaily) · 2026-10-27 02:52

Rock Matrix 开源 EdgeRock-4B，将 Qwen3.5-4B 的 256K KV 缓存从 8 GiB 削减至 72 MiB

Built from Qwen3.5-4B with HKUST and CMU, EdgeRock-4B keeps a fixed 72 MiB KV cache at 256K context and serves 6.42 times the throughput while keeping most of its reasoning scores.