Meridian State UniversityResearch Computing
Restricted Internal
AI Model Card · Internal Summary
Open-weight LLM (Large) — Local
| Document type | Internal Model Summary | Version | 1.1 |
| Runs on | Secure Research AI Cluster | Visibility | Restricted Internal |
| Hosting | On-prem (open weights) | Vendor card | None — internal only |
| Last reviewed | 2026-03-22 | Reviewer | Research Computing |
1. Model details
Model family
Large open-weight language model, self-hosted.
Weights
Pinned version; stored and run entirely within the cluster.
Connectivity
No outbound inference calls; fully offline to the open internet.
2. Intended use
Text analysis on sensitive or regulated research corpora that cannot leave the institutional boundary. Approved for use only within an active, reviewed research project.
3. Out-of-scope use
- General-purpose chat or productivity tasks (use Meridian GPT)
- Any processing outside the approved project's data-use agreement
- Redistribution of weights or outputs beyond the project team
4. Evaluation & limits
Local validation
Validated on a representative sample before each study's production use.
Known limits
Quality below frontier hosted models; can hallucinate — outputs require expert review.
5. Local deployment notes
Model version is pinned per project for reproducibility. Updates are evaluated and documented before a project is moved to a new version. Prompts and outputs are retained within the project workspace only.
Review & approval
Model reviewer — Research Computing
Principal investigator
AI Governance Committee
Date approved