OpenAI's ChatGPT voice mode runs on an older, weaker GPT-4o era model with an April 2024 knowledge cutoff, despite users expecting it to be the most advanced interface. This reflects a growing capability gap between different AI access points, with enterprise-focused models receiving more development resources due to verifiable reward functions and business value. The disparity highlights how AI performance varies significantly across different domains and deployment methods.
Background
OpenAI offers multiple AI interfaces with varying capabilities, where voice mode appears to use older technology compared to their most advanced text-based models. This reflects broader industry trends where different AI applications receive unequal development focus based on business priorities.
- Source
- Simon Willison
- Published
- Apr 10, 2026 at 11:56 PM
- Score
- 6.0 / 10