Questions
- How do I choose the right AI model in LibreChat?
- What are the differences between the available AI models?
- When should I use an Agent instead of a model?
- What types of Agents are available in LibreChat?
- How do I access and use Agents within LibreChat?
Environment
This information applies to Bowdoin College faculty, staff, and students using LibreChat through a web browser on Bowdoin-managed or personal devices. Users must log in with Bowdoin Single Sign-On and multi-factor authentication. Model and Agent availability may change as new options are added and old ones are retired.
Resolution
LibreChat offers a secure interface for prompting a variety of generative AI models along with the ability to use your documents to augment AI models' capabilities. It can serve as a centralized hub for all your AI conversations, providing a familiar, user-friendly interface enriched with advanced features and customization capabilities.
Choosing an AI model or Agent in LibreChat
In LibreChat, choosing the right AI model depends on the complexity, purpose, and sensitivity of your task. Different models may vary in speed, cost, reasoning ability, and strengths—such as summarizing text, generating creative content, or answering technical questions. When selecting a model, consider the nature of your request (e.g., factual accuracy vs. creative output), the size or depth of the content you need, and any data privacy or compliance requirements set by your institution. Consulting LibreChat’s model descriptions (see below) and experimenting with smaller test prompts can help you identify which model delivers the most suitable and reliable results for your academic or administrative needs.
LibreChat has a wide variety of generative AI models for you to choose from. The current list will be updated as new models become available and older models are deprecated. At this time the default model is GTP-5.
Comparison of current LibreChat AI models
| Model Group & Examples |
General Strengths |
Typical Use Cases |
Considerations |
|
OpenAI GPT‑5 Series
(GPT‑5, GPT‑5 Chat Latest, GPT‑5 Mini, GPT‑5 Nano)
|
Latest and most capable OpenAI reasoning models with strong creativity and accuracy; different sizes trade speed vs. depth. |
Complex academic writing, multi‑step reasoning, generating instructional content. |
Larger models are slower/more expensive; smaller versions (Mini, Nano) are faster but less nuanced. |
|
OpenAI GPT‑4o Series
(GPT‑4o, GPT‑4o Mini)
|
Balanced speed and intelligence; good at multimedia/multi‑modal reasoning (if enabled). |
Day‑to‑day question answering, short essays, tutoring-style explanations. |
Smaller versions save resources with slightly less depth. |
|
Other OpenAI Lightweight Models
(01, 03‑mini, 01‑mini)
|
Very fast, low‑cost, good for simple or repetitive queries. |
Quick fact lookups, drafting emails, short instructional prompts. |
Limited reasoning and creativity compared to flagship models. |
|
Anthropic Claude (via AWS Bedrock)
Claude 3.5 Sonnet, Claude 3.7 Sonnet, Claude Sonnet 4, Claude Opus 4
|
Strong at reasoning, summarizing, and following complex instructions in a clear, balanced style. Often excels at nuanced writing with a broad knowledge base. |
Research summaries, academic writing assistance, policy drafting, sensitive-topic handling. |
Generally slightly slower than smaller models; choose latest version for updated knowledge and reasoning. |
|
Amazon Nova
(Pro v1.0))
|
Optimized for accurate, fast, reliable responses; good for structured outputs. |
Administrative workflows, data extraction, academic Q&A. |
May be less creative than Claude or GPT for storytelling; Premier may have higher cost or latency. |
|
Meta LLaMA
(LLaMA 3 & 4 Maverick)
|
Good open‑source models for general use, especially when cost efficiency is important. |
Brainstorming, summarizing readings, producing drafts where slightly less polish is acceptable. |
May be less nuanced than top-tier proprietary models in very complex tasks. |
|
DeepSeek
(R1)
|
High‑speed reasoning and code generation; good at technical problem-solving. |
Programming help, data analysis, technical troubleshooting. |
May produce more concise than elaborate prose—best for technical or math-heavy needs. |
Choosing an Agent in LibreChat
In LibreChat, an Agent is a specialized AI assistant that operates with a defined purpose, set of tools, and instructions to accomplish specific types of tasks. Unlike a general chat session, an Agent can be configured with custom knowledge or integrate with external data sources and follow persistent behaviour rules to handle complex or recurring workflows. These Agents have been pre-configured and shared with all users. Currently you can choose from the Image Generation Agent (generates images using OpenAI's model) and the Artifacts Agent (Use LibreChat's Artifacts feature for generating diagrams, preview generated HTML, React, and more!).
Agents can be accessed from the AI Models drop down menu in the top left of the main chat window:
Image Generation Agent example:
If you have questions, please contact the Service Desk at (207) 725-3030 or servicedesk@bowdoin.edu