CatChat
Overview
CatChat is a campus-wide generative AI service provided by the Research Cyberinfrastructure (RCI) team at Montana State University. It gives faculty, staff, and students a single, secure login to a curated set of large language models, including Google Gemma, Alibaba Qwen, OpenAI o4-mini, Meta Muse, and Mistral AI Mistral. Every model is self-hosted on MSU's Tempest cluster, so your prompts and responses never leave campus infrastructure.
Catchat is free to use for all Students / Faculty / Staff at Montana State University and approved for use with data classified as “Public” (data which may be released to the general public in a controlled manner) or “Low / Medium” (such as FERPA protected data, financial transactions which don’t include confidential data, contracts, etc.) by the Bozeman Data Stewardship Standards.
The above approval for public and Low / Medium data types does not imply approval
for any specific use cases. It is important to follow institutional, departmental,
and instructor guidelines for allowable and appropriate uses of these tools. Catchat
and the models contained within are experimental and may produce inaccurate or incomplete
results so output should always be verified.
Environmental Impact
Catchat is maintained by Research Computing with computational resources provided by the Tempest supercomputer which is powered by approximately 80% renewable energy and cooled with systems that do not consume water.
Getting Started
How to Use
-
Access and Security
-
Model Selection
-
Tool Usage
-
Agent Creation
-
Connect to External Providers
-
API Access
-
Connect to IDE
-
Connect to Claude Desktop
Available Models
| gemma4 (Google) (gemma4-31b) |
qwen3.8 (Alibaba) (qwen3.8-27b) |
o4-mini (OpenAI) (gpt-oss) |
muse-glimmer (Meta) ( muse ) |
mistral-small:24b (MistralAI) ( mistral ) |
Use Cases
|
Use Case |
Requirements |
Recommended Models |
|---|---|---|
|
Advanced reasoning and multi-domain research |
Highest accuracy, long-context reasoning, strong factual recall, and cross-domain synthesis (science, engineering, policy, etc.) |
o4-mini, qwen3.8 |
|
Enterprise copilots / knowledge assistants |
Strong reasoning, safe output, high reliability, API latency tolerance |
o4-mini, gemma4 |
|
General-purpose chatbots and customer service |
Balanced cost–performance, multilingual fluency, context awareness |
gemma4, mistral |
|
Code generation and debugging (multi-language) |
Syntax awareness, structured reasoning, efficiency on moderate hardware |
o4-mini, qwen3.8, mistral |
|
Lightweight internal automation (summaries, tagging, extraction) |
Low latency, scalable inference, moderate reasoning depth |
muse, mistral
|
|
Fine-tuning / domain adaptation experiments |
Model that trains efficiently, flexible licensing |
gemma4, mistral |
|
Creative writing, marketing copy, dialogue |
Natural tone generation, stylistic diversity, moderate context memory |
gemma4, qwen3.8 |
|
Technical document summarization |
Structured understanding, moderate token context, reliability |
qwen3.8, o4-mini |
|
Research prototyping / evaluation of LLM benchmarks |
Open weights, transparency, large context |
qwen3.8, o4-mini, mistral |
Usage/Data Privacy
CatChat is run locally at MSU so your data never leaves the university or becomes public. Data in CatChat is private by default, with student data protected by FERPA regulations. More information is given in the Bozeman Data Stweardship Standards.
|
|
Public Data |
Low Risk Data |
Medium Risk Data |
High Risk Data |
Specialized Risk Data |
| Public AI Tools (Personal OpenAI, Claude..etc) |
Yes |
No |
No |
No |
No |
| Microsoft CoPilot when signed in with MSU Account |
Yes |
Yes |
Yes |
No |
No |
| CatChat |
Yes |
Yes |
Yes |
No |
No |
Other tools and requests must be reviewed by the MSU software procument process. Some are in progress. Contact uitprocurement@montana.edu to request software review. Full data storage security information can be found here.