Quality measures whether a language-model response is legible, useful, accurate, and appropriate for a specific task. Hallucination risk matters alongside response usefulness. section 2.5.1
Assess quality by:
- Defining the problem and intended use before choosing a model. section 2.5.1
- Trying different models with comparable prompts and examining their outputs. section 2.5.1
- Balancing response quality against practical factors such as model size, cost, and platform capabilities. section 2.5.1
Code-generation comparison
Experimentation may show that OpenAI GPT models produce better code examples than Google Gemini for a particular task; the result should guide a deliberate choice rather than a general ranking. section 2.5.1
