ChatGPT Voice Mode Revealed Running on GPT-4o-Era Model — 13-Month Capability Gap With Text Interface

BuildFastWithAI·medium signal

Developer investigation revealed that ChatGPT's Voice Mode operates on a GPT-4o-era model with an April 2024 knowledge cutoff while the text interface uses GPT-5.5/GPT-5.5 Instant — a 13+ month capability gap between modalities. The discovery sparked backlash from Andrej Karpathy and Simon Willison, with OpenAI citing real-time latency constraints as the reason. For builders relying on voice-driven AI workflows, this means voice interactions are substantially less capable than text despite appearing to use the same product.