AI News | 2 min read

OpenAI Opens GPT-Live-1.1 Voice Model to Developers With 60% Cheaper Idle Pricing

OpenAI has released GPT-Live-1.1 — the full-duplex voice model powering ChatGPT — to external developers via API, with 38 voice IDs and a new standby tier that cuts idle costs by 60%.

Hector Herrera
Hector Herrera
A newsroom where a person is building related to a major AI company Opens a technology company-Live-1.1 Voice
Why this matters OpenAI has released GPT-Live-1.1 — the full-duplex voice model powering ChatGPT — to external developers via API, with 38 voice IDs and a new standby tier that cuts idle costs by 60%.

OpenAI Opens GPT-Live-1.1 Voice Model to Developers With 60% Cheaper Idle Pricing

By Hector Herrera | November 20, 2026

OpenAI has released GPT-Live-1.1 — the full-duplex voice model that has powered ChatGPT Plus and Pro since October — to external developers via API. The release includes 38 voice IDs, a new standby pricing tier at $0.02 per minute, and improved multilingual support, giving builders the tools to ship real-time voice applications without building the voice stack from scratch.

Background

Full-duplex voice means the model can listen and speak simultaneously — the way a real conversation works, rather than the push-to-talk pattern that dominated voice interfaces for years. OpenAI introduced this capability to ChatGPT subscribers in October 2026. Opening it to developers via API follows the company's pattern of shipping to its own products first, then unlocking the underlying model for the developer ecosystem. GPT-Live-1.1 is an incremental update over the original GPT-Live model, with improved pronunciation accuracy and accent consistency.

What's in the Release

The developer release ships several concrete changes:

  • 38 voice IDs — including seven new options with Spanish-leaning phonology, targeting the large Spanish-speaking developer and user base
  • Standby mode — billed at $0.02 per minute when the connection is open but no active speech is happening, versus $0.05 per minute for active sessions; that's a 60% reduction in idle costs
  • Improved multilingual transcription — the model handles code-switching (switching between languages mid-sentence) more reliably than the prior version
  • Pronunciation and accent consistency — voices hold accent and intonation more stably across long sessions, reducing the "drift" that made earlier voice models sound inconsistent over time

Why Idle Pricing Matters

The standby pricing tier is the most practically significant change for developers building real-world voice products. Voice interfaces often involve long pauses — a user thinking, background noise, or simply a conversation lull. Under previous pricing, every second of open connection cost the same regardless of activity. For applications like voice-enabled customer service, tutoring, or accessibility tools, those idle seconds add up fast. The new two-tier structure makes the economics significantly more predictable for production deployments.

Impact

For voice application developers: The API access removes a major barrier. Teams building voice-first products no longer need to assemble their own stack of speech-to-text, language model, and text-to-speech components. A single API call handles the full loop in real time.

For multilingual products: The seven new Spanish-leaning voices are a targeted expansion. Spanish is the second most-spoken language globally by native speakers, and voice AI adoption in Latin America and Spain has lagged partly due to accent inconsistency in earlier models. This release directly addresses that gap.

For OpenAI's competitive position: Google's Gemini Live and Anthropic's voice capabilities are both moving fast. Releasing GPT-Live-1.1 to developers keeps OpenAI's voice stack in the hands of the developer community that builds the products end users actually use — a distribution advantage that matters as much as raw model quality.

What to Watch

The practical test will be latency at scale. Full-duplex voice over API introduces round-trip delay that ChatGPT manages with infrastructure optimizations not always available to external developers. Watch the developer community forum over the next few weeks for real-world latency numbers from teams building production apps. If idle pricing makes voice products economically viable at scale, expect a wave of new voice-first applications in early 2027.

Key Takeaways

  • ✓ By Hector Herrera | November 20, 2026
  • ✓ 60% reduction in idle costs
  • ✓ Improved multilingual transcription
  • ✓ Pronunciation and accent consistency
  • ✓ For voice application developers:

Did this help you understand AI better?

Your feedback helps us write more useful content.

Hector Herrera

Written by

Hector Herrera

Hector Herrera is an AI systems architect in Houston and founder of Hex AI Systems. He designs and runs AI systems in production and writes daily about how AI is reshaping business, government and everyday life. 20+ years building for the web. Houston, TX.

More from Hector →

Get tomorrow's AI briefing

Join readers who start their day with NexChron. Free, daily, no spam.

More from NexChron