MemT 1.0 Models by MemTensor
MemTensor releases · 2 in the last 12 monthsChangelog →
2 models2026
Last refreshed 2026-10-20. Next refresh: weekly.
About
MemT 1.0 is MemTensor's family of memory-native language models, released October 20, 2026. A native memory block runs alongside the backbone's attention so past turns are stored in a fixed-size memory state instead of being reloaded as context. Nano is built on Qwen3.6-35B-A3B (Apache 2.0) and Flash on DeepSeek-V4-Flash (MIT).
Decision facts
- Best fit
- agent workflows
- Capability starting point
- MemT-1.0-Flash
- Lowest tracked input
- Not tracked
- Closest related family
- Agents-A1
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
2 in view
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| MemT-1.0-Nano | Use when provider availability and model metadata match the workload. | 2026-10 | — | Current |
| MemT-1.0-Flash | Use when provider availability and model metadata match the workload. | 2026-10 | — | Current |
Release Timeline
1 release group2026-10
2 current
Current
Current
Specifications(2 models)
| Model | Released | Parameters |
|---|---|---|
| MemT-1.0-Nano | 2026-10 | 36B (3B activated) |
| MemT-1.0-Flash | 2026-10 | 288B (17B activated) |