OpenAI has released a major update to its programming interfaces (API), making interactions with artificial intelligence faster and more accurate. The main changes focused on real-time AI assistants and voice-processing models.

Error-free voice communication

The new gpt-realtime-1.5 model has significantly improved speech recognition reliability. One key improvement is a 10% increase in transcription accuracy for numbers and letters, which is especially important when dictating contact information, addresses, or passwords.

AI now understands complex instructions 7% better. Voice assistants will make fewer mistakes and execute user requests more accurately on the first attempt, making interactions with the neural network more natural.

Instant response speed

The second major update concerns the speed of “smart” agents. With the introduction of the new data transmission technology (WebSocket), the system no longer needs to reprocess all information each time. A persistent connection now transmits only updates.

According to OpenAI, this has increased the performance of complex AI agents by 20–40%. For regular users, this means near-instant responses: delays when executing commands have almost disappeared.

ORIENT