Select the right model
There are multiple models on the campusGenAI platform to choose from. When selecting a model, consider:
- The Complexity of the Task: Larger models like GPT-5, Deepseek-R1, and LLaMA 3 70B provide stronger reasoning capabilities that are best suited for large-scale and complex tasks.
- Speed Requirements: Smaller models like Claude 4.5 Haiku and GPT-5 mini are optimized for real-time applications.
- Provider Preferences: If you have specific security or compliance needs, you can choose models from preferred providers, including AWS, Azure, and OpenAI.
Web search and image generation
Certain models have been integrated with web search and/or image generation tools.
- Models with the Magnifying Glass emoji (🔍) are integrated with web search. You can prompt the model to query the Internet and use its findings in its response. It is important to be explicit about asking the model to do this, otherwise it will rely on the data it was trained on, which may not contain up-to-date information on current events.
- Models with the Framed Picture emoji (🖼) are integrated with image generation. You can prompt the model to generate an image by providing a brief description of the desired image.
Data cutoff and web search
Most AI models are trained only up to a fixed point in time. For example, OpenAI's newest model GPT-5 is trained only up until September 2024. This is not something that we (the campusGenAI platform team) can control—it is the same regardless of where you access the model, and only changes when OpenAI, Anthropic, or another provider releases a newer model trained on more recent data.
That said, we have recently added web search and image generation to our most popular models. With web search, you can explicitly ask the model to query the internet for current information, which it can then use in its response, regardless of its training cutoff.
