tags:
- base-model
- mistral
- unsloth
- local-ai
- sovereign-ai
- foundation
- text-generation
base_model: mistralai/Mistral-Small-24B-Instruct-2501
🏗️ Mistral-Small-24B-Instruct-2501 (Unsloth Base Model)
The Architectural Foundation of the EleMo Ecosystem
This repository hosts the unmodified base model
Mistral-Small-24B-Instruct-2501, optimized via
Unsloth.
We provide this model on the KI-Insel (AI Island) repository hub for two primary reasons:
- Absolute Transparency: To allow researchers, educators, and institutions to see the exact foundational architecture we use to train our specialized pedagogical models.
- Guaranteed Availability: To ensure our underlying infrastructure remains permanently accessible for our own fine-tuning pipelines and community reproducibility.
⚠️ Important Disclaimer: This is NOT EleMo
Please be aware that this is the raw, general-purpose language model.
It does not contain the pedagogical fine-tuning, the Margaret Carr methodological guardrails, or the specific tone adjustments developed by Kita Digital. If you use this model out-of-the-box for pedagogical documentation, it will likely hallucinate, make inappropriate clinical diagnostics, or fail to adhere to early childhood education standards.
- Looking for the pedagogically trained model?
Please visit our official, fine-tuned Stable-Releases here: EleMo- Stable-Realeses or explore our Adapter Collection.
⚙️ Technical Details & Local Execution
We chose the Mistral-Small-24B architecture as our foundation because it strikes the perfect balance between highly capable reasoning and local deployability. It is large enough to understand complex, unstructured observational data, yet efficient enough to run locally without cloud dependency.
Ideal for Data Sovereignty:
This model can be executed fully offline on local high-performance hardware (e.g., Apple Silicon with high Unified Memory or dedicated local AI workstations), fully supporting our Zero-Cloud Guarantee.
How to use this base model (via LM Studio)
If you wish to test the raw capabilities of this model or use it to load our EleMo LoRA adapters:
- Download: Navigate to the "Files and versions" tab and download the appropriate
.gguf file for your system's VRAM/RAM capacity.
- Load: Open LM Studio and drag the file into the interface.
- Run Locally: Ensure your network connection is disabled if you are testing with sensitive data, verifying true local execution.
- Apply Adapters: You can now navigate to the "Fine-tuning / Adapters" section in LM Studio to lay the
EleMo-LoRA weights on top of this foundation.
Hosted by Kita Digital - Building the infrastructure for Sovereign AI in early childhood education.
