Test-Time Compute
Test-time compute is the practice of allocating additional computational resources during inference — when the model generates a response — rather than during training. Reasoning models like OpenAI's o1/o3 and DeepSeek-R1 use test-time compute to "think" through problems step by step, exploring multiple approaches and evaluating potential solutions before producing a final answer.
Example
Given a complex math olympiad problem, a standard model responds in under a second but gets it wrong. A reasoning model with test-time compute spends 30 seconds generating internal chain-of-thought reasoning — testing approaches, catching errors, and reconsidering — before delivering a correct answer, trading speed for accuracy.
Put this into practice
Build polished, copy-ready prompts in under 60 seconds with SurePrompts.
Try SurePrompts