OpenAI bets on voice as AI’s primary interface with new models, and crypto builders should pay attention
Three new real-time voice models signal a platform shift that could reshape how users interact with everything from DeFi protocols to on-chain agents.
OpenAI dropped three new voice AI models on May 7: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. Together they handle real-time conversation, live translation across more than 70 languages, and low-latency transcription.
What the new models actually do
GPT-Realtime-2 is the flagship of the bunch, sitting within OpenAI’s GPT-5 tier. It packs a 128K context window, which means it can hold the equivalent of a short novel’s worth of conversation in memory during a single session.
The model scored a 15.2% improvement on the Big Bench Audio benchmark compared to its predecessors.
GPT-Realtime-2 supports parallel tool calls, adjustable reasoning effort, and what OpenAI describes as “strong recovery behavior.”
Pricing is set at $32 per million audio input tokens and $64 per million output tokens for the core model. Translation runs $0.034 per minute, transcription $0.017 per minute.
Early adopters already in production include Zillow, which is using voice agents for home searches, Deutsche Telekom for multilingual customer support, and Priceline for conversational trip management.
The Jony Ive connection and the hardware play
OpenAI acquired Jony Ive’s design firm for $6.5 billion in May 2025, a deal that signaled the company’s ambitions extend well beyond software. The combination of world-class industrial design and voice-first AI models points toward a hardware product, likely a personal device that treats voice as the default input method.
Why crypto builders should care
A voice-first paradigm could collapse friction dramatically for DeFi applications. Projects building AI agents that interact with blockchains, from wallet management to yield optimization, should be watching OpenAI’s pricing and API access closely. At $32 per million input tokens, the economics of wrapping a voice layer around an on-chain agent are entering viable territory for consumer products, not just enterprise deployments.
Investors in the AI-crypto intersection should watch whether decentralized compute networks and inference marketplaces begin supporting real-time audio workloads. The latency requirements for voice are unforgiving, and most decentralized infrastructure wasn’t built with that constraint in mind.