Why the Model Is the Wrong Starting Point for AI Apps

When it comes to choosing models, developers often start with a frontier model, which is another name for a cutting-edge, large language model such as OpenAI’s GPT-4, Google’s Gemini Ultra, and Anthropic’s Claude 3 series.

But frontier models can quickly become cost-prohibitive. That’s when developers tend to find smaller models that may trade some accuracy for costs, according to Forrester principle analyst Rowan Curran, whose research focuses on AI, machine learning and data science.

After seeing what can be accomplished with a big…