The L&S AI Inference platform provides a selection of models optimized for different tasks. We recommend the following flagship models for most work.
Best for: Everyday questions, writing assistance, summarization, brainstorming, drafting, Q&A, and describing images
Strengths:
Limitations:
Use it when: You want a quick, clear assistant for text-based or image-based tasks.
Best for: Long documents, scientific/STEM research, complex document analysis, and code
Strengths:
Limitations:
gemma-4-26b-a4b-itUse it when: You need to process very long documents, work with scientific or mathematical material, or write and debug code.
Best for: Complex analysis, multi-step problems, research synthesis, and demanding logic
Strengths:
Limitations:
Use it when: The task genuinely requires heavy-duty reasoning or deep synthesis.
Additional models may be available through the API for specific use cases, but they do not receive full documentation support and may change without notice. See API Access for how to list the models currently available.
| Task | Recommended model | Why |
|---|---|---|
| Summarize a 2-page text | gemma-4-26b-a4b-it | Fast and well suited to short-form work |
| Summarize a 500-page book | qwen3.8-27b | 256k context window handles large volumes |
| Complex research synthesis | gpt-oss-120b | Best for deep reasoning across many sources |
| Task | Recommended model | Why |
|---|---|---|
| Write a Python script | qwen3.8-27b | Strong code generation |
| Debug a function | qwen3.8-27b | Strong at multi-step reasoning and tool use |
| Architect a system | gpt-oss-120b | High-level reasoning across complex constraints |
| Task | Recommended model | Why |
|---|---|---|
| Describe a screenshot | gemma-4-26b-a4b-it | Fast image understanding |
| Analyze a complex chart | qwen3.8-27b | Stronger reasoning over visual content |
Questions or issues? Contact help@cit.ucsb.edu.