ChatGPT Supercharges Long-Form Chat with 94% Faster Loading: 741-turn Conversation Reduced from 28 Seconds to 2 Seconds

According to Dynamic Beating monitoring, OpenAI is preparing to optimize the ultra-long conversation experience of ChatGPT and Codex, set to be launched next week. After an internal test of a massive 741-turn, 231MB chat, the average load time has decreased from 27.62 seconds to 1.66 seconds, resulting in a speed improvement of about 94%.

This optimization primarily focuses on chat history loading, rather than model response time. During testing, the number of chat entries to load decreased from 15,529 to 64, the number of requests dropped from 894 to 16, and the overall app memory growth also reduced by 41.2%.

For users who regularly use ChatGPT or Codex, this will significantly reduce the waiting time when opening old conversations. In the past, the longer the thread, the laggier it became as the client had to handle a large number of historical messages; this time OpenAI directly reduced the amount of data that needs to be loaded.