Temperature Compare
See how temperature affects language model output. Compare the same prompt at different temperature settings side-by-side to understand the tradeoff between predictability and creativity.
Log in to use this tool.
Pedagogical Goals
- •Make the abstract concept of temperature concrete and visual
- •Show that temperature controls randomness, not creativity or intelligence
- •Let students discover that low temperature produces repetitive but safe output, while high temperature produces varied but sometimes nonsensical output
- •Build intuition for why different applications need different temperature settings
How It Works
The tool sends the same prompt to GPT at two different temperature settings simultaneously and streams both responses side-by-side. Students can adjust the temperatures and try different prompts to see how the outputs diverge. The streaming display makes the generation process visible in real time.
How It Was Built
Built as a client component with dual streaming responses. The API endpoint accepts a temperature parameter and forwards to Azure OpenAI. The side-by-side layout uses a split panel with synchronized streaming, so students can watch both completions emerge simultaneously.