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

T2Covered by 1 source · first reported by China: SCMP Tech / TechNode / Pandaily / KrASIA / Sixth Tone at 27 Oct, 02:52 UTC

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.

Sources

• T2Rock Matrix Open-Sources EdgeRock-4B, Cutting Qwen3.5-4B's 256K KV Cache From 8 GiB to 72 MiBChina: SCMP Tech / TechNode / Pandaily / KrASIA / Sixth Tone27 Oct, 02:52