Meridian State UniversityOffice of the CIO · AI Services
Restricted Internal
AI Model Card · Internal Summary
General Assistant (Large)
| Document type | Internal Model Summary | Version | 1.3 |
| Offered via | Meridian GPT | Visibility | Restricted Internal |
| Provider | Contracted vendor (hosted) | Vendor card | Public — linked |
| Last reviewed | 2026-05-10 | Reviewer | AI Services + Governance |
1. Model details
Model family
Large general-purpose language model (vendor-hosted).
Modality
Text in / text out; long-context capable.
Hosting
Vendor API via the MSU AI Gateway; not hosted on campus.
2. Intended use
Selected as the default model in Meridian GPT for reasoning-heavy tasks: analyzing long documents, drafting and editing, and coding assistance. Appropriate for Public and Internal data within the gateway's controls.
3. Out-of-scope & prohibited use
- Processing Confidential or regulated data
- Automated decisions about people without human review
- Any use outside the MSU AI Gateway's approved controls
4. Training data
Trained by the vendor on a broad corpus; specifics are documented in the vendor's public model card. MSU does not fine-tune this model, and university data is not used in training.
5. Evaluation & performance
Local evaluation
Spot-checked on representative campus tasks (summarization, Q&A, coding) before enablement.
Result
Strong on reasoning and long context; selected as default for those tasks.
Benchmarks
Vendor-reported benchmarks referenced in the public model card.
6. Limitations & known risks
- Can be confidently wrong (hallucination); outputs require verification
- Knowledge cutoff applies; not a source of current facts
- May reflect biases present in training data
7. Local deployment notes
Routed as the default model for new conversations in Meridian GPT. If the vendor deprecates or materially changes this model, AI Services re-evaluates and updates this summary before changing the default.
8. Ethical considerations
Use is governed by the institution's AI principles: human-centered, transparent, private and secure, and equitable. Accessibility of the Meridian GPT client is evaluated separately in the supporting evidence layer.
Review & approval
Model reviewer — AI Services
AI Governance Committee
Vendor card reviewed (link on file)
Date approved