AI Models

what llms and machine learning models work for your ai solution

This card covers which language or machine learning models fit your tool. Different models vary in cost, accuracy, language coverage, and where they run. The choice is practical, not about picking the most advanced option, and it shapes what the tool can and cannot do.

Questions to explore

// use these as prompts in a workshop or on your own. There are no right answers.

What does the task actually require from a model, and does a simpler one suffice?
How well does the model handle the languages and topics your newsroom works in?
Where does the model run, and what does that mean for your data and sources?
How will you compare a few models on your own real examples before choosing?
What happens to your tool if the model you picked changes, gets pricier, or shuts down?

Expert voices

// notes from the journalists and AI experts who helped shape this kit

“How do you adjust the model: train a new one, fine-tune an existing one, or prompt an unadjusted model to complete your task? These are real alternatives worth discussing separately.”

Peter Deselaers, DW Akademie

“Even AI scientists cannot always explain why a model produced a given output. You are working with black boxes, which makes transparency about how you use them essential.”

Walid Al-Saqaf, Doha Institute for Graduate Studies

“The focus is on large language models, but what could we build with small language models or simple machine learning?”

Carl Javier, Data and AI Ethicist

“Put GPT, Claude, Gemini, and others side by side on the same task. Comparing models directly teaches more about their limits than any single-tool tutorial.”

Steffen Leidel, DW Akademie

Things to consider

The newest or largest model is not always the right fit for the job.
Test models on your own newsroom examples, not on generic benchmarks.
Note that model behavior can shift over time without notice.

using this card

Pull AI Models when it is relevant and set it aside when it is not. Pair it with the other AI Solutions cards, lay them out on a table, and use the questions above to get everyone on the same page. Capture what you discuss on sticky notes or in a shared doc.