Available AI Models

📌 Purpose

This article provides guidance for selecting an AI model in AI Salve. Each model has different strengths, including everyday writing, research, complex reasoning, coding, document analysis, and privacy-focused local processing.

Selecting the right model can improve the quality and speed of a response. When choosing a model, consider the complexity of the request, whether current web information is needed, and whether the information should remain within Salve Regina University's locally hosted AI environment.

👥 Audience

• Faculty
• Staff
• Students

🖥️ Applies To

• Salve AI
• Web-based applications
• Cloud-hosted AI models
• Locally hosted AI models

🛠️ Choosing an AI Model

✅ Recommended Selection Process

1. Identify whether the task is simple, professional, technical, research-oriented, or highly complex.
2. Determine whether local processing is preferred for the information involved.
3. Choose the model whose recommended use cases most closely match the task.
4. Enter a clear prompt that includes the goal, relevant context, and preferred output format.
5. Review the response for accuracy, completeness, and appropriate source attribution.
6. Try a different model if the first response does not meet the needs of the task.

Use the following recommendations to select the model that best matches your task.

Mistral Small

Best for fast, straightforward tasks.

• Answering routine questions
• Summarizing shorter content
• Rewriting or improving text
• Brainstorming ideas
• Classifying or organizing information
• Assisting with simple coding tasks

Choose Mistral Small when speed and efficiency are more important than advanced reasoning. For complicated analysis or difficult technical problems, consider GPT 6, Claude Opus 5, or Claude Sonnet 5.

Claude Opus 5

Best for difficult tasks, in-depth analysis, and advanced reasoning.

• Analyzing complicated problems
• Working with long or detailed documents
• Producing polished long-form writing
• Handling sophisticated coding and software architecture tasks
• Comparing multiple perspectives or requirements
• Completing work where response quality is more important than speed

Choose Claude Opus 5 for high-complexity assignments that require careful reasoning, detailed analysis, or high-quality writing.

Grok 4.6

Best for research-oriented and technical questions.

• Exploring technical topics
• Solving reasoning problems
• Assisting with programming and troubleshooting
• Comparing technologies or approaches
• Researching complicated subjects
• Working with current information when supported web-search tools are enabled

Choose Grok 4.6 when the task emphasizes research, technical exploration, or problem solving. Always review cited sources when a response relies on web search.

GPT 6

Best as a general-purpose cloud model for a wide variety of tasks.

• Everyday questions and writing assistance
• Advanced reasoning and analysis
• Coding and technical troubleshooting
• Reviewing documents and files
• Research-oriented workflows
• Tasks involving multiple tools or steps
• Working across different types of content

Choose GPT 6 when you are unsure which cloud model to use or when a task combines writing, analysis, coding, files, and research.

Claude Sonnet 5

Best for everyday professional work, document analysis, and coding.

• Drafting and revising professional content
• Summarizing documents
• Analyzing written information
• Writing, reviewing, and explaining code
• Completing general reasoning tasks
• Balancing response quality with speed

Choose Claude Sonnet 5 for routine professional and academic work that still requires strong reasoning. For especially difficult or lengthy tasks, consider Claude Opus 5.

Gemma 4 (Local)

Best for private, locally processed, everyday AI tasks.

• General questions and conversations
• Summarizing internal content
• Drafting and rewriting text
• Working with approved internal knowledge sources
• Knowledge-base and retrieval-augmented generation workflows
• Tasks where local processing is preferred

Gemma 4 is hosted within Salve's local AI environment. Choose it for everyday work when local processing is preferred over a cloud-hosted model.

NVIDIA Nemotron (Local)

Best for private technical work and reasoning tasks.

• Technical questions and explanations
• Coding assistance
• Reasoning and problem solving
• Internal knowledge-base workflows
• Retrieval-augmented generation workflows
• Experiments involving tools or agent-style workflows
• Tasks where local processing is preferred

NVIDIA Nemotron is hosted within Salve's local AI environment. Choose it when a locally processed task requires more emphasis on reasoning, coding, or technical content.

☁️ Cloud Models vs. Local Models

Cloud models may provide access to more advanced capabilities, but requests are processed through an external AI provider. Examples in Salve AI may include:

• Mistral Small
• Claude Opus 5
• Grok 4.6
• GPT 6
• Claude Sonnet 5

Local models run within Salve's locally hosted AI environment. Examples include:

• Gemma 4
• NVIDIA Nemotron

💡 Troubleshooting & Tips

• Start with GPT 6 when you are unsure which cloud model to select.
• Use Mistral Small for quick and straightforward requests.
• Use Claude Sonnet 5 for everyday writing, document analysis, and coding.
• Use Claude Opus 5 when a task requires deeper analysis or more advanced reasoning.
• Use Grok 4.6 for technical exploration and research-oriented questions.
• Use Gemma 4 Local for private, general-purpose work .
• Use NVIDIA Nemotron Local for private technical, coding, and reasoning tasks.
• Do not assume that a confident response is accurate. Verify important facts using authoritative sources.
• For web-assisted answers, open and review the sources provided by the model.
• AI model names, availability, and capabilities may change as Salve AI is updated.