| language: en | |
| license: apache-2.0 | |
| tags: | |
| - mtrinido | |
| - moe-experts | |
| - wnmoe | |
| - manifold | |
| model_name: Mtrini-Tellus 128B Expert Manifold | |
| # 🧠 Mtrini-Tellus 128B Expert Manifold | |
| This repository contains the **dynamic weight shards** for the Mtrini-Tellus 128B architecture. | |
| ### ⚡ Role in WNMOE | |
| These experts are not loaded at once. The **WNMOE (When Needed Mixture of Experts)** system uses this repository as a remote hydration source, pulling individual expert tensors into the active compute manifold based on token-routing requirements. | |
| - **Expert Count**: 128 Clusters | |
| - **Architecture**: Neuron-Packed Sparse Manifold | |
| - **Verification**: Compiwer AI Environment Standard | |
| *This manifold is managed by the Mtrinido V4 AI Factory.* |