On-premises large language model access for UCSB students, staff, and faculty. All inference runs on UCSB hardware — your data never leaves campus.
| Chat interface | chat.ai.college.ucsb.edu |
| API endpoint | api.ai.college.ucsb.edu/v1 |
| Sign in | UCSB NetID and Password |
| API keys | By request — email [contact] |
| Status | Early Access (Alpha/Beta) |
Be sure to check out the Frequently Asked Questions, the Getting Started Guide, and the list of known limitations.
Open WebUI User Guide Model switching, document upload and RAG, web search, system prompts, and tips for getting better results from open-weight models.
Choosing the Right Model When to use gemma-4-31b, qwen3-coder-next, and gpt-oss-120b — with task-by-task examples.
Context Windows and Effective Prompting How context limits work, strategies for long conversations, and how prompting open-weight models differs from frontier commercial models.
Working configurations validated by CIT staff for some popular coding agents.
opencode Terminal-based AI coding agent — full provider config and agent role setup.
aider AI pair programming in the terminal — environment variable and YAML config, architect mode.
Continue (VS Code) Open-source VS Code extension — config.yaml setup, model roles, autocomplete, and secure key storage.
Pi Coding Agent Lightweight, minimal, "power user" coding agent with "less is more" mentality paired with "everything is a typescript extension" mentality.
| Model | Best for | Context |
|---|---|---|
| gemma-4-31b | General chat, writing, summarization | 128k |
| qwen3-coder-next | Code generation, debugging, review | 128k |
| gpt-oss-120b | Complex reasoning, long documents, research synthesis | 128k |
| granite‑4.1‑30b | General use, strong language comprehension, fast | 128k |
Questions, issues, or feedback: help@cit.ucsb.edu
This is an early access service. Bug reports and feature requests are welcome and help improve the platform.