How Tolan builds voice-first AI with GPT-5.1

Source: OpenAI Date: 2026-01-07 URL: https://openai.com/index/tolan

Summary

OpenAI case study from January 2026 featuring Tolan — a company building voice-first AI applications on top of GPT-5.1. Tolan’s use case illustrates the real-time audio API capabilities introduced with GPT-4o’s voice mode and extended through the GPT-5.x family, enabling low-latency conversational AI with natural speech. Published as GPT-5.1 became the recommended API model for new deployments, this piece served as evidence of voice-first architectures becoming production-viable.

Implications

Voice as a first-class API modality. Tolan’s product illustrates the shift from voice as a ChatGPT feature to voice as a platform capability developers could build on. The GPT Realtime API (launched October 2024) made this possible; GPT-5.1’s improvements in conversational coherence and latency made it compelling.

Thread: Voice-first applications. Sits alongside the Voice Engine safety documentation, the GPT-4o voice mode launch, and the GPT Realtime API as the progression from “text with TTS” to “native real-time conversation” as an application paradigm.

Watch: Whether Tolan and similar voice-first products achieved meaningful adoption, and how latency and quality continued to improve through the GPT-5.x cadence.