Jump to content

LibreChat: Difference between revisions

From Swarthmore Knowledge Base
← Older edit
VisualWikitext
No edit summary
 
Line 32: Line 32:
| variant = info
| variant = info
}}
}}
{| class="wikitable"
{| class="wikitable sortable"
!Model
! Model Name !! Suggested Use !! Quota Guidance
!Suggested end-user description
!Quota guidance
|-
|-
|'''claude-sonnet-5'''
| claude-fable-5 || Frontier flagship for the most demanding/complex tasks || Premium
|'''Recommended for most work.''' Strong for drafting and revising writing, summarizing readings, developing course materials, planning projects, analyzing documents, and programming help.
|'''Standard'''
|-
|-
|'''claude-fable-5'''
| claude-opus-5 || Current Anthropic flagship for complex reasoning || High
|'''For exceptionally difficult work.''' Use for complex research questions, advanced analysis, major coding projects, or long, multi-step tasks when other models have not produced a satisfactory result.
|'''Premium — use selectively'''
|-
|-
|'''claude-opus-4-8'''
| claude-sonnet-5 || General workhorse — best Claude price/performance || Medium-High
|Best for careful analysis, nuanced writing, complex research, and difficult programming. Consider Sonnet first for routine work.
|'''Premium'''
|-
|-
|'''claude-haiku-4-5'''
| claude-opus-4-8 || Older-gen flagship|| High
|'''Fast and economical.''' Good for short summaries, email drafts, rewriting, brainstorming, classification, and straightforward questions. Start here for simple tasks.
|'''Low'''
|-
|-
|'''gpt-5.5'''
| claude-haiku-4-5 || Fast, cheap Claude tier for lightweight tasks || Medium
|A powerful model for complicated professional and academic work, including document analysis, data interpretation, research planning, and advanced programming.
|'''Premium'''
|-
|-
|'''gpt-5.6-terra'''
| gpt-5.6-sol || Flagship agentic coding || High
|'''Recommended general-purpose GPT model.''' Good for writing, teaching materials, research assistance, document analysis, data work, and moderate programming.
|'''Standard'''
|-
|-
|'''gpt-5.6-luna'''
| gpt-5.6-terra || Balanced general tier || Unknown / Verify
|'''Fastest and most economical GPT option.''' Good for quick questions, editing, summaries, brainstorming, routine emails, and simple explanations.
|'''Low'''
|-
|-
|'''gpt-5.6-sol'''
| gpt-5.6-luna || High-volume frontier-class model || Low
|'''For the most demanding tasks.''' Best for difficult reasoning, advanced research, complex data analysis, and large programming projects. Try Terra before using Sol.
|'''Premium — use selectively'''
|-
|-
|'''gpt-oss-20b'''
| grok-4.3 || General use || Low-Medium
|An economical open model for routine questions, basic writing, technical explanations, and programming help. Results may require more checking and revision than premium models.
|'''Low'''
|-
|-
|'''deepseek-v3.2'''
| gemma-4-31b || Current-gen dense flagship, general use || Very Low
|A cost-conscious choice for mathematics, scientific and technical explanations, reasoning, data analysis, and programming. Useful when a stronger reasoning model is needed without using a premium option.
|'''Low–Standard'''
|-
|-
|'''qwen3-coder-30b'''
| gemma-4-e2b || Tiny edge-class model — cheapest in catalog || Minimal
|'''Economical programming assistant.''' Good for explaining code, debugging assignments or research scripts, generating examples, and working with Python, R, JavaScript, and other languages.
|'''Low'''
|-
|-
|'''qwen3-coder-480b'''
| deepseek-v3.2 || General/reasoning tasks || Low
|For difficult programming and software-development tasks, especially large codebases, complex debugging, or multi-file projects. Excessive for routine coding questions.
|'''Premium'''
|-
|-
|'''qwen3-coder-next'''
| qwen3-next-80b || Fast general use || Low
|A strong coding-focused option for developing, debugging, testing, and documenting larger projects. A good middle ground between the 30B and 480B coding models.
|'''Standard'''
|-
|-
|'''grok-4.3'''
| qwen3-235b || Primary general-purpose Qwen model, cheapest output || Low
|A versatile model for writing, analyzing long documents, working with images, producing structured information, and handling multi-step tasks.
|'''Standard'''
|-
|-
|'''glm-5'''
| qwen3-coder-next || Coding-focused tasks || Low
|Best for complex programming, technical analysis, research workflows, and projects that require many connected steps. Use GLM-4.7 for less demanding work.
|'''Premium'''
|-
|-
|'''glm-4.7'''
| glm-5 || Flagship agentic coding || Low-Medium
|A versatile option for writing, research assistance, presentations, creative work, and programming. Suitable as a general-purpose alternative to Claude or GPT.
|'''Standard'''
|-
|-
|'''glm-4.7-flash'''
| glm-4.7-flash || Ultra-cheap, high-volume tasks || Minimal
|'''Fast and economical.''' Good for summaries, translations, quick drafts, simple coding questions, and routine administrative work.
|'''Low'''
|-
|-
|'''kimi-k2.5'''
| kimi-k2.5 || General use with built-in reasoning || Low-Medium
|Particularly useful when your prompt includes '''screenshots, diagrams, charts, scanned pages, or interface designs'''. Also capable of writing, reasoning, and programming.
|'''Standard'''
|-
|-
|'''kimi-k2-thinking'''
| minimax-m2.5 || Agentic/coding-focused tasks || Low
|For difficult academic reasoning, mathematics, research synthesis, complex programming, and problems that require sustained step-by-step analysis.
|'''Premium'''
|-
|-
|'''minimax-m2.5'''
| mistral-large-3-675b || Mistral flagship, general use || Low
|A cost-effective option for office and academic productivity, including documents, spreadsheets, presentations, research tasks, and programming.
|-
|'''Low–Standard'''
| devstral-2-123b || Dedicated coding tasks || Low-Medium
|-
| ministral-3-14b-instruct || Best small Ministral model, symmetric I/O pricing || Minimal
|-
| voxtral-small-24b-2507 || Audio-input tasks (only audio model in list) || Minimal
|-
| nemotron-super-3-120b || Current Nemotron flagship, general use || Very Low
|-
| nemotron-nano-3-30b || Small model, excellent price/performance || Minimal
|}
|}


Latest revision as of 13:52, 2 September 2026

Artificial intelligence chatbot service

Faculty, Staff, Students

Web


No

Property Value
Description Artificial intelligence chatbot service
Access Faculty, Staff, Students
Platform Web


Description

LibreChat is an online service that provides a secure interface to a variety of Artificial Intelligence (AI) language models, including ChatGPT and Claude.

All the models used are private and none of them train on data entered to the site.

Getting Started

  1. Must be on campus or using SwatVPN
  2. Go to librechat.ai.swarthmore.edu
  3. Log in with your Swarthmore Google account
  4. Orienting to the screen (see the graphic below for a visual representation):
    1. On the upper left you will find a dropdown of the different AI companies
    2. To the right of that, another dropdown allows you to select the specific AI model you’d like to experiment with.
    3. At the bottom of the page is where you enter your prompt. A prompt can be entered as text or you can speak your prompt using the microphone at the far right of the prompt field.
  5. Start experimenting by entering text in the chat box
LibreChat interface showing dropdown menus to select AI company and model, a chat prompt asking How can I help you today?, and a text box for entering messages or uploading files
e

Models

LibreChat provides several different AI models from leading companies as well as open-weights (non-proprietary) options. This list changes frequently. The list below shows options as of Summer 2026.

Model Name Suggested Use Quota Guidance
claude-fable-5 Frontier flagship for the most demanding/complex tasks Premium
claude-opus-5 Current Anthropic flagship for complex reasoning High
claude-sonnet-5 General workhorse — best Claude price/performance Medium-High
claude-opus-4-8 Older-gen flagship High
claude-haiku-4-5 Fast, cheap Claude tier for lightweight tasks Medium
gpt-5.6-sol Flagship agentic coding High
gpt-5.6-terra Balanced general tier Unknown / Verify
gpt-5.6-luna High-volume frontier-class model Low
grok-4.3 General use Low-Medium
gemma-4-31b Current-gen dense flagship, general use Very Low
gemma-4-e2b Tiny edge-class model — cheapest in catalog Minimal
deepseek-v3.2 General/reasoning tasks Low
qwen3-next-80b Fast general use Low
qwen3-235b Primary general-purpose Qwen model, cheapest output Low
qwen3-coder-next Coding-focused tasks Low
glm-5 Flagship agentic coding Low-Medium
glm-4.7-flash Ultra-cheap, high-volume tasks Minimal
kimi-k2.5 General use with built-in reasoning Low-Medium
minimax-m2.5 Agentic/coding-focused tasks Low
mistral-large-3-675b Mistral flagship, general use Low
devstral-2-123b Dedicated coding tasks Low-Medium
ministral-3-14b-instruct Best small Ministral model, symmetric I/O pricing Minimal
voxtral-small-24b-2507 Audio-input tasks (only audio model in list) Minimal
nemotron-super-3-120b Current Nemotron flagship, general use Very Low
nemotron-nano-3-30b Small model, excellent price/performance Minimal

Use Cases and Ideas

  • Ask the AI tool to generate ideas for a title of a blog post or project
  • Copy and paste some text and ask to refine or simplify
  • Upload a file and ask to generate a summary.

Other Notes

  • There are additional features such as Agents, Assistants, and Plugins that aren't fully configured by ITS yet and may not work.
  • LibreChat tools can create diagrams and provide code but cannot generate files such as image files or PDFs. You can upload files and the AI can interpret them, but it cannot generate new files.

Troubleshooting

File Uploads

Some AI models can only handle uploads using certain file types. If you get an error when trying to upload a file, start a new chat, then select the "Upload as Text" option, which converts the contents of the file to text before sending it to the AI model.

Retrieved from "https://kb.swarthmore.edu/index.php?title=LibreChat&oldid=4070"