OpenAI releases faster gpt-realtime-2.1 voice models

OpenAI released two new Realtime API models on July 6 - gpt-realtime-2.1 and gpt-realtime-2.1-mini - aimed at building low-latency voice and multimodal agents. The company said improved caching cut 95th-percentile latency by at least 25% across its Realtime voice models, while the full model adds better alphanumeric recognition, silence and noise handling, and interruption behavior. The mini tier brings realtime reasoning and tool use at a lower price point, framing the update as production-focused rather than a new frontier model.