OpenAI Adds Reasoning Effort Slider for ChatGPT Plus and Pro Users
OpenAI has rolled out a new UI slider that allows ChatGPT Plus and Pro users to manually adjust the amount of reasoning effort the model applies to each response. This feature gives users direct control over the balance between response speed and depth of thought. This update improves user experience by letting users save time and compute resources on simple queries, while allocating more inference-time compute for complex tasks. It marks a shift towards giving users granular control over LLM resource consumption and latency. The slider targets OpenAI's reasoning models, which rely on scaling inference-time compute to solve difficult problems. While not a major technological breakthrough, it provides a practical UX solution to manage the latency inherent in deep-thinking models.
## BACKGROUND
Traditional LLMs generate answers quickly based on pre-trained patterns, but newer reasoning models (like OpenAI's o-series) use "inference-time compute" to think step-by-step before responding. This process improves accuracy for complex math, coding, and logic tasks but increases latency and computational cost. Allowing users to adjust this effort helps balance speed and quality depending on the task.