staccs/lecore-qwen35-9b-assimilated
The staccs/lecore-qwen35-9b-assimilated model is a 9 billion parameter Qwen3.5-9B checkpoint that has undergone leCore's Unicron assimilation process. This process integrates the model with holographic memory (HRR) capabilities, enabling damage-tolerant memory and extended context handling up to 262,144 native tokens. It is designed for local-first deployments requiring robust long-context processing and composable memory operations, with a focus on memory retention rather than a capability increase over the base model.
Loading preview...
Model Overview
staccs/lecore-qwen35-9b-assimilated is a 9 billion parameter model based on the Qwen3.5-9B architecture, processed through leCore's Unicron assimilation gate. This integration focuses on enhancing the model's memory capabilities, particularly through holographic memory (HRR), rather than improving its core language model performance. The assimilation process ensures the model's weights are honest and wired towards this advanced memory system.
Key Capabilities & Features
- Extended Context Window: Offers a native context window of 262,144 tokens, significantly larger than many standard chat models, suitable for processing entire books or large codebases. Upstream also supports YaRN extension up to ~1M tokens.
- Damage-Tolerant Memory (HRR): Utilizes holographic memory, which is resilient to corruption. Unlike traditional memory that fails catastrophically, HRR allows for graceful degradation, maintaining recall even with significant data loss.
- Composable Memory Operations: Supports "bind/unbind" operations for composing roles and facts using invertible math (FHRR), enabling flexible attribute-value associations.
- HoloMachine: Allows for running short, inspectable, and deterministic holographic instruction streams directly on the HRR core, avoiding black-box weights.
- Memory Optimization: While the 9B model's perplexity remains unchanged, the assimilation process has been shown in other runs (e.g., 72B) to significantly reduce model size (e.g., 145GB to 57.8GB for 72B) with minimal perplexity impact, making it suitable for resource-constrained environments.
Use Cases & Differentiators
This model is particularly well-suited for applications requiring robust long-context understanding and resilient memory management. Its primary differentiator is the integration with leCore's holographic memory, which provides a unique approach to handling large amounts of information and maintaining state. It's ideal for local-first deployments where memory efficiency and fault tolerance are critical, offering a path into the leCore ecosystem for advanced memory operations. It's important to note that this is a "prove run" demonstrating the assimilation process and HRR surface, not a claim of improved core language model capabilities over the base Qwen3.5-9B.