OpenAI shipped three new voice models to its Realtime API on May 7, 2026. GPT-Realtime-2 is the first voice model with GPT-5-class reasoning, scoring 15.2% higher on Big Bench Audio and 13.8% higher on Audio MultiChallenge than GPT-Realtime-1.5; it supports configurable reasoning effort and tool use for complex voice-agent workflows. GPT-Realtime-Translate handles live, in-call translation across 70+ input languages into 13 output languages in a single streaming pass. GPT-Realtime-Whisper delivers partial transcript deltas as the speaker talks, priced at $0.017 per minute. The Realtime API simultaneously exits beta and reaches general availability.