OpenAI enhances voice agents with GPT-5 class reasoning

OpenAI has introduced three new voice models, including GPT-Realtime-2, which features GPT-5 class reasoning for more natural and complex conversations. These models are designed to reduce the overhead typically associated with voice agents by separating tasks like conversation, translation, and transcription into specialized components. This allows enterprises to route discrete voice tasks to appropriate models, improving efficiency and orchestration. The new offerings compete with similar solutions from Mistral and aim to capitalize on the growing enterprise interest in voice agents due to increased user comfort and valuable interaction data.