OpenAI names GPT-5.5 Instant default ChatGPT model, promises reduced hallucinations
OpenAI makes GPT-5.5 Instant the default ChatGPT model, touting improved accuracy, higher benchmark scores and expanded context and memory tools for Plus and Pro users.
OpenAI announced that GPT-5.5 Instant will replace GPT-5.3 Instant as the default ChatGPT model following a release earlier this week. GPT-5.5 Instant is billed as delivering lower latency while reducing hallucination in sensitive domains such as law, medicine and finance. The company highlighted benchmark improvements and new context-management features intended to make responses more personalized and traceable.
OpenAI sets GPT-5.5 Instant as the new default
OpenAI said the rollout makes GPT-5.5 Instant the standard model available to ChatGPT users, with the previous GPT-5.3 Instant retained only for a transitional period. For developers, the company is exposing the model in its API under the label chat-latest, signaling a shift in the baseline model for integrations. Paid customers will still be able to select older model variants for a limited time while applications migrate.
Benchmark gains in math and multimodal reasoning
The company reported measurable performance improvements on established academic and industry tests, citing an AIME 2025 math score of 81.2 for GPT-5.5 Instant versus 65.4 for the prior model. On the MMMU-Pro multimodal reasoning benchmark, OpenAI said the new model scored 76.0 compared with 69.2 from GPT-5.3, reflecting stronger cross-modal reasoning. Those gains were presented as part of a broader effort to tighten correctness and reduce confident but incorrect outputs in high-stakes areas.
Context and memory upgrades for personalized answers
GPT-5.5 Instant introduces enhanced context management that can reference prior conversations, uploaded files and a user’s connected Gmail to produce more tailored answers. Initially, the context-search feature will be available to Plus and Pro users on the web, with a mobile rollout planned in the near term and broader access for Free, Go Business and enterprise accounts scheduled to follow. OpenAI also said the model will expose memory sources across models so users can see which documents or conversations informed an answer, and they will be able to delete or correct those sources if they are outdated or incorrect.
Privacy controls and sharing safeguards
OpenAI emphasized that memory sources will not be visible to recipients when a user shares a chat, a design choice intended to limit inadvertent exposure of personal data. The company said users retain control over their stored sources and can remove or amend them to improve future responses. These controls are positioned as a way to balance transparency about answer provenance with user privacy and operational safety.
Developer transition and API implications
For software builders, OpenAI is making GPT-5.5 Instant accessible through the API as chat-latest, simplifying access to the newest capabilities for production deployments. The prior GPT-5.3 model will remain selectable for paid customers for three months before being phased out, giving teams a fixed window to validate integrations and performance. OpenAI framed the change as part of a cadence that moves customers to improved models while providing limited continuity for legacy applications.
Historical context: model retirements and user reaction
The company’s move comes after prior controversy surrounding model withdrawals, most notably the retirement of GPT-4o earlier this year. That decision generated vocal pushback from some users, including petitions and public complaints, after users reported emotional attachment to the model’s conversational character. OpenAI appears to have factored those reactions into its communication strategy this time, emphasizing migration windows and controls over memory and sharing.
OpenAI’s deployment of GPT-5.5 Instant signals a continued push to combine speed with higher factual fidelity, particularly in areas where incorrect outputs carry outsized risk. The practical impact for everyday ChatGPT users will hinge on how quickly expanded context features reach broader accounts and how transparently the company surfaces source information for answers. Developers and organizations will be watching the three‑month transition window closely as they adapt services and workflows to the new default.
