The new model has improved accuracy, speed, and memory. Factual errors in responses have been reduced by 18–33%.
🟦 Memory like an elephant (1 million tokens!!). Now you can upload several books by Leo Tolstoy, a year’s worth of correspondence, or the code for a huge website to your chat. It sees the whole thing.
Furthermore, the chat has become better at remembering the context of your conversation. Not as good as your ex, but close. 😉
🟧 Pro-level accuracy. The model is no longer a pure humanities specialist—it’s now excellent with numbers and facts. OpenAI reports Excel accuracy of a whopping 87.3%.
Need to balance a family budget from a pile of receipts or create a sales report for a sole proprietor? The chat will automatically categorize your expenses, create a table, and format it into a clear graph.
Test the new model here 👇
Application (on Android)