OpenAI Rolls Out GPT-5.5 Instant as ChatGPT's New Default Model
GPT-5.5 Instant Replaces GPT-5.3 Instant
OpenAI rolled out GPT-5.5 Instant as ChatGPT's default model for all users — including the free tier — on May 5. In the API, the model is accessible via the chat-latest endpoint, replacing GPT-5.3 Instant.
The new default delivers sharper, more concise answers and improved personalization drawn from past conversations, uploaded files, and Gmail integrations. OpenAI emphasized better reasoning capabilities and accuracy as the headline improvements over its predecessor.
Three New Real-Time Voice Models in the API
Alongside the default model change, OpenAI added three specialized voice models to the API: GPT-Realtime-2 for live voice reasoning, GPT-Realtime-Translate for real-time multilingual speech, and GPT-Realtime-Whisper for streaming transcription.
The full GPT-5.5 was released to paid subscribers on April 23; the Instant version extending access to free users marks another step in OpenAI's strategy to push its latest-generation models as broadly as possible.
Sources: OpenAI Blog · TechCrunch
Related Articles
OpenAI가 ChatGPT 기본 모델을 GPT-5.5 Instant으로 교체했다. 의료·법률·금융 등 고위험 분야에서 환각 오류가 52.5% 줄었고, 응답 길이도 30.2% 단축됐다.
ChatGPT의 유료 경험이 Instant와 deep reasoning 사이를 오가던 구조에서 GPT-5.6 Sol 중심으로 정리된다. OpenAI는 고위험 사실성 평가에서 GPT-5.5 Instant보다 사실 오류 응답이 68% 적었다고 밝혔다.
OpenAI가 GPT-5.3-Codex를 공개하며 SWE-Bench Pro 등 실제 소프트웨어 엔지니어링 벤치마크에서 SOTA 달성. 개발 과정에서 초기 버전이 자신의 학습을 디버깅하고 배포를 관리했다.