OpenAI Introduces GPT-Realtime-2: GPT-5-Class Reasoning for Voice Agents

Original: OpenAI Introduces GPT-Realtime-2: GPT-5-Class Reasoning for Voice Agents View original →

Read in other languages: 한국어日本語
AI May 8, 2026 By Insights AI (Twitter) 1 min read 2 views Source

Real-Time Voice Reasoning Arrives

OpenAI has launched GPT-Realtime-2 in its API — the company's first voice model powered by GPT-5-class reasoning. The model enables voice agents to listen, reason, and take action as conversations unfold, moving beyond simple call-and-response interactions.

New Streaming Models

Alongside GPT-Realtime-2, OpenAI released two additional streaming models:

  • GPT-Realtime-Translate: Live speech translation from 70+ input languages into 13 output languages
  • GPT-Realtime-Whisper: Streaming speech-to-text that transcribes as you speak

Shift in Voice AI Capabilities

The new models represent a significant upgrade to what voice agents can accomplish. GPT-Realtime-2 can handle interruptions, maintain context across long conversations, and reason through complex requests — all in real time. The context window has expanded from 32K to 128K tokens compared to the previous Realtime API.

Pricing

GPT-Realtime-2 is priced at $32 per million audio input tokens and is available immediately via the API.

Share: Long

Related Articles

Comments (0)

No comments yet. Be the first to comment!

Leave a Comment