OpenAI has released two significant API updates aimed at enhancing developer functionality. The new gpt-realtime-1.5 model improves voice command reliability, showing a ten percent boost in accurate transcription of numbers and letters, a five percent increase in logical audio processing, and a seven percent enhancement in instruction adherence during internal testing. Additionally, the audio model has been upgraded to version 1.5.
The Responses API now supports WebSockets, enabling a persistent connection that only transmits new data rather than the full context for each request. This enhancement accelerates the performance of complex AI agents with multiple tool interactions by 20 to 40 percent.
For more insights into AI developments, subscribe to THE DECODER for ad-free content, a weekly AI newsletter, and exclusive reports. Stay updated on the latest advancements in AI technology and improve your development experience with these updates.
Source link
