OpenAI Releases GPT-Realtime-2.1 and ...
Quick Answer
OpenAI has launched GPT-Realtime-2.1 and GPT-Realtime-2.1-mini, enhancing low-latency voice agents with integrated reasoning capabilities at the same cost as previous models.
Quick Take
The mini version offers improved performance with a 25% reduction in p95 latency and significantly lower audio output costs, making reasoning the default feature in voice applications.
Key Points
- GPT-Realtime-2.1-mini offers reasoning at the same price as the old model, $0.60/1M text.
- P95 latency reduced by at least 25% due to improved caching across voice models.
- Audio output for mini is $20/1M, compared to $64/1M for the full model.
- Reasoning effort can be adjusted from minimal to high, optimizing latency and depth.
- Full model enhances alphanumeric recognition and noise handling capabilities.
📖 Reader Mode
~1 min readOpenAI Releases GPT-Realtime-2.1 and GPT-Realtime-2.1-mini for Low-Latency Voice Agents in the API For years, the cheap voice model was the "fast but not smart" tier. You picked reasoning or you picked low cost — never both. OpenAI just deleted that trade-off. They shipped two new Realtime models: gpt-realtime-2.1 and gpt-realtime-2.1-mini. The mini is the one worth your attention. It's a mini reasoning model for realtime voice — and it lands at the exact same price as the old gpt-realtime-mini. Here's what's actually interesting: → Reasoning and tool use now run on the mini tier, at the same cost as gpt-realtime-mini — $0.60/1M text in, $10/$20 audio in/out → Reasoning effort is a dial: minimal → xhigh, low by default — throttle latency against depth on every turn → p95 latency is down at least 25% across Realtime voice models, purely from improved caching → Audio output runs $20/1M on the mini vs $64/1M on the full 2.1 — roughly 3x cheaper, reasoning intact → The full gpt-realtime-2.1 adds sharper alphanumeric recognition, better silence and noise handling, and cleaner interruption behavior The main effort is to make reasoning the default in voice, not the paid upgrade. Full analysis: marktechpost.com/2026/07/06/ope… Technical details: developers.openai.com/api/docs/guide…
@OpenAI@OpenAINewsroom@OpenAIDevs— Originally published at x.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from WebSearch (Tavily)
See more →全球AI芯片峰会，9月上海见！
The 2026 Global AI Chip Summit will take place in Shanghai on September 22-23, focusing on the evolving AI chip landscape, including the shift from training to inference, the rise of diverse chip technologies, and the restructuring of industry competition. Notable speakers include experts from leading universities and companies, discussing advancements in AI chip architecture and applications.