MemTensor Open-Sources MemT 1.0, Memory-Native Models That Keep Long Histories in a Fixed State

T2Covered by 1 source · first reported by China: SCMP Tech / TechNode / Pandaily / KrASIA / Sixth Tone at 22 Oct, 01:50 UTC

MemTensor released MemT 1.0 Flash and Nano, models with a built-in trainable memory state. It reports about 4x faster inference at 256K history and 130x smaller storage.

aireleases

Sources

• T2MemTensor Open-Sources MemT 1.0, Memory-Native Models That Keep Long Histories in a Fixed StateChina: SCMP Tech / TechNode / Pandaily / KrASIA / Sixth Tone22 Oct, 01:50