To get started with AI Commons, see the helpful resources on this page including answers to common questions, explanation of tokens, prompting basics and more.
Get Started
1. Sign In
Use your UCSBNetID single sign-on.
2. Choose a model
Pick the right AI model for your task.
3. Start a chat
Ask, paste text, or upload compatible documents.
4. Explore features
Create a bot, build an API, or get support.
UCSB AI Commons is a secure, enterprise-grade generative AI platform engineered for architectural versatility. Designed to support diverse institutional workflows-ranging from standard conversational interfaces and custom Retrieval-Augmented Generation (RAG) agents to robust API integrations-the platform provides a controlled environment for scalable innovation. Hosted within the UCSB AWS Cloud, AI Commons adheres strictly to NIST SP 800-171 compliance standards, ensuring rigorous data protection and governance across all platform operations.
AI Commons is currently available to students in AI CoP, staff, and faculty with an active university account.
Undergraduate students do not currently have access to AI Commons.
AI Commons can be used for everyday work and research tasks such as drafting and editing text, summarizing documents, brainstorming ideas, answering questions, and writing or explaining code. custom bots, file upload, API access upon approved request. The platform is suitable only for non-sensitive institutional data.
The main difference is that AI Commons is provided and managed by UCSB, while public AI tools are outside services with their own terms.
In AI Commons:
- Prompts and outputs are processed within a NIST 800-171 compliant environment on AWS, protected by strict access controls, identity verification and service control policies.
- Access requires a UCSBnetID to log in, and use of this platform is limited to eligible university users.
- No personal subscriptions. Each user receives a standard quota of 1 million tokens per month.
- Users can chat with leading models from multiple providers-currently spanning Anthropic's Claude family (Haiku, Sonnet, and Opus), Meta Llama 4, Amazon Nova, Qwen3, and OpenAI's open-weight GPT-OSS models.
Choosing the Right Tool
Comparing AI Commons vs. Google Gemini for Education
| Feature/Use Case | UCSB AI COmmons | Google Gemini For education |
|---|---|---|
| Best for... | Comparing different AI models, technical experimentation, and building custom bots/APIs. | Everyday productivity, student data, and seamless integration with Gmail and Google Docs. |
| Model Variety | Yes. Test and compare multiple foundational models. | No. Exclusive to Google’s AI models. |
| Custom Bots & APIs | Yes. Includes API access and custom bot creation tools. | No. Designed as a ready-to-use application. |
| Data Security & PII | Secure for specific regulated data (NIST 800-171), but requires advance approval for Controlled Unclassified Information (CUI). | Covered by existing UCSB data protection; ideal for general institutional data and student PII. |
| Ease of Use | Moderate. Requires some familiarity with token budgets and model types. | High. Intuitive and seamlessly built into your daily Google Workspace. |
Choose UCSB AI Commons
If you want to compare models, build a custom bot, or
need API access
Choose Gemini
If you want simple AI assistance right inside your daily Google apps or
are handling standard student data
Explore
Learn the essentials
Tap any card to flip it and read a quick explainer.
Model Selection Guide
Compare the Claude family, Meta Llama 4, Amazon Nova, Qwen3, and OpenAI's models. See the Models page for full comparisons.
Model Selection Guide
Pick the right model for your task and speed.
Tokens Explained
Input, output, and monthly usage.
Tokens Explained
Tokens are pieces of text an AI chatbot processes when reading your prompt and generating a response. Tokens can be words, parts of words, etc. Both input tokens (your prompts) and output tokens (chatbot response) count toward your monthly token usage. The model you choose affects how many tokens are consumed and the associated cost.
Multimodal Capabilities
Upload diagrams, screenshots, or charts. Do not upload identifiable people or sensitive documents.
Multimodal Capabilities
Work with images and text together in one conversation.
Prompting Basics
Write clear instructions to get better, more reliable answers.
Prompting Basics
State the role, the task, the audience, and the format. Give examples. Ask for revisions.
Knowledge Bases (RAG)
Upload approved PDFs, Markdown, or text. The bot retrieves relevant passages and cites them.
Knowledge Bases (RAG)
Ground responses in your own documents and sources.
Embedding Models
Turn text into vectors to power search and retrieval.
Embedding Models
Embeddings let bots find the most relevant passage in a knowledge base. Managed automatically for RAG bots.