OpenAI Launches GPT-5.5 Instant with 50% Higher Accuracy and Total Memory Control

In March, OpenAI unveiled the GPT-5.3 Instant model, which exhibited a more organic conversational cadence and a diminished propensity for refusal. OpenAI has now introduced its successor: GPT-5.5 Instant. This nascent model is slated to become the default engine for all ChatGPT users, including those on the complimentary tier.

The most profound enhancement within GPT-5.5 Instant lies in its factual veracity. According to OpenAI’s internal benchmarks, the model has achieved a 52.5% reduction in hallucinations across the medical, legal, and financial domains compared to its predecessor. Furthermore, owing to advancements in visual reasoning, mathematics, and scientific proficiency, the model’s performance in image analysis and STEM-related inquiries has been significantly bolstered. Illustratively, it achieved scores of 85.6% on the GPQA, 81.2% on the AIME 2026, and 76% on the MMMU-Pro assessments.

OpenAI maintains that GPT-5.5 Instant will provide more precise, intelligent, and succinct responses. Consequently, this should mitigate the verbosity and repetitive phrasing often associated with AI dialogues, delivering accurate answers with brevity and sparing users from exhaustive reading or redundant follow-up inquiries.

Beyond the introduction of GPT-5.5 Instant, OpenAI announced that all ChatGPT models will now support memory attribution. When ChatGPT provides a personalized response, users will be presented with the specific contextual metadata utilized—such as stored memories or excerpts from historical conversations—and will possess the autonomy to edit or expunge obsolete information.

GPT-5.5 Instant is currently being deployed to ChatGPT Free, Go, Plus, and Pro subscribers. Within the API ecosystem, it is designated as chat-latest, permitting developers to invoke the model programmatically. As per standard protocol, all premium subscribers will retain access to GPT-5.3 Instant for a three-month transitional period prior to its eventual decommissioning.

The refined personalization features—which aggregate insights from historical dialogues, uploaded documents, and integrated Gmail data—will initially be available to ChatGPT Plus and Pro users on the web interface. This functionality is expected to debut on mobile platforms shortly, with an expansion to Free, Business, and Enterprise tiers anticipated in the coming weeks.