{"name":"MemT-1.0","repo":"MemTensor/MemT-1.0","url":"https://midorreal.com/project/memtensor-memt-1-0","repository":"https://github.com/MemTensor/MemT-1.0","description":"Official repository for MemT-1.0","facts":{"language":{"value":"Python","source":"https://github.com/MemTensor/MemT-1.0","checked":"22 Nov 2026"},"releases":{"value":"0","source":"https://github.com/MemTensor/MemT-1.0/releases","checked":"20 Nov 2026"},"contributors":{"value":"4","source":"https://github.com/MemTensor/MemT-1.0/graphs/contributors","checked":"20 Nov 2026"},"archived":{"value":"no","source":"https://github.com/MemTensor/MemT-1.0","checked":"22 Nov 2026"}},"stars":545.0,"writeup":{"what_it_is":"MemT-1.0 is a family of memory foundation models that adds persistent, dynamically evolving memory to model computation. The repository contains the official code, including a SGLang-based inference server and a Python client for conversations with native memory.","audience":"Researchers and developers working on language model memory and agentic coding.","claims":[{"kind":"specific","claim":"Persistent state stays constant in size, reducing storage by about 66 times versus a full-context KV cache at 128K history length.","excerpt":"Its persistent state remains constant in size, reducing storage by approximately 66× versus full-context KV cache at 128K history length.","status":"not_checked"},{"kind":"specific","claim":"Trained on 34B tokens of data with a two-stage training pipeline.","excerpt":"supported by 34B tokens of data and dedicated infrastructure","status":"not_checked"},{"kind":"specific","claim":"Across 11 memory benchmarks, the Pro, Flash and Nano models rank first, second and third among evaluated parametric-memory methods.","excerpt":"Across 11 memory benchmarks, MemT-1.0-Pro, MemT-1.0-Flash, and MemT-1.0-Nano rank first, second, and third among the **evaluated parametric-memory methods**, respectively.","status":"not_checked"},{"kind":"specific","claim":"Retrieved experience is stored in native memory and accessed during execution, without adding experience text to the task context.","excerpt":"Retrieved experience is stored in native memory and accessed through memory attention during execution","status":"not_checked"},{"kind":"specific","claim":"Memory state can be saved and restored, so a session can resume without the previous conversation transcript.","excerpt":"Restore memory without the previous conversation transcript.","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-21"},"why_now":null,"owner_supplied":null,"score":{"verdict":"mid","hype":33,"reality":51,"gap":-19,"momentum":null,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T08:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-10-19T03:29:30+00:00","text":"Repository created","source":"https://github.com/MemTensor/MemT-1.0"}],"cite":"MemT-1.0 is a GitHub project at https://github.com/MemTensor/MemT-1.0 written mainly in Python, described there as \"Official repository for MemT-1.0\".","attribution":"Data from Mid or Real, https://midorreal.com/project/memtensor-memt-1-0, checked 22 November 2026. Attribution with a link is required.","still_real":null}