OpenAI Rolls Out GPT-5.5 Instant as ChatGPT Default, Replacing GPT-5.3 Instant

OpenAI has rolled out GPT-5.5 Instant as the default model in ChatGPT, replacing GPT-5.3 Instant for all users, with the company framing the change as an accuracy upgrade for everyday use.

The New Stack reports that OpenAI is replacing the default model in ChatGPT with GPT-5.5 Instant and says it “produces faster responses, shorter answers, and fewer incorrect claims.”

9to5Google

The New Stack also ties the update to earlier Instant releases, noting that GPT-5.5 Instant follows GPT-5.3 Instant, introduced in March, and sits alongside the flagship GPT-5.5 model introduced in April.

TestingCatalog similarly says OpenAI is rolling out GPT-5.5 Instant as the new default model in ChatGPT, replacing GPT-5.3 Instant for all users, and describes it as targeting ChatGPT’s “highest-volume everyday model.”

Axios adds that OpenAI updated ChatGPT’s default model “to respond with more accuracy, more personalization and fewer gratuitous emoji,” and says GPT-5.5 Instant is rolling out to all ChatGPT users and to the API.

Multiple outlets also describe the rollout as immediate for the default, while paid users keep GPT-5.3 Instant for a limited overlap period: Mashable says GPT-5.3 Instant “will remain available for the next three months for paid users but will otherwise be sunsetted,” and TestingCatalog likewise says paid users will keep access to GPT-5.3 Instant for three months before it is retired.

In parallel, OpenAI is positioning GPT-5.5 Instant as the everyday “daily driver for hundreds of millions of people,” a framing Seeking Alpha attributes to OpenAI’s comments about why “small improvements make” a difference for the default experience.

OpenAI’s central claim for GPT-5.5 Instant is that it reduces hallucinations and improves factuality, with several outlets repeating the same internal evaluation figures.

The New Stack says OpenAI states in a blog post that “accuracy” is the model’s core selling point, and it describes GPT-5.5 Instant as “Instant is now more dependable, with significant improvements in factuality across the board and the largest gains in domains where accuracy matters most.”

Breakingthenews.net

Mashable reports that OpenAI says GPT-5.5 Instant produced “52.5 percent fewer hallucinated claims” in internal testing than GPT-5.3 in “high stakes” topics like law, finance, and medicine, and it adds that the new model “reduced inaccurate claims by 37.3% on especially challenging conversations users had flagged for factual errors.”

TestingCatalog repeats the same two figures—“52.5% fewer hallucinated claims” and “37.3%”—and adds that the update brings stronger performance across “visual reasoning, math, science, STEM questions, uploaded image analysis, and web-search decision-making.”

The Decoder provides a broader set of benchmark comparisons, including that on AIME 2025 accuracy jumped from 65.4 to 81.2 percent, and it also cites GPQA rising from 78.5 to 85.6 percent, CharXiv going from 75.0 to 81.6 percent, and MMMU-Pro rising from 69.2 to 76.0 percent.

The Decoder also includes an additional test metric, saying “The error rate on OmniDocBench… dropped from 14.6 to 12.5 percent,” which it presents as part of the same “tighter answers and smarter personalization” effort.

Decrypt ties the launch to a specific example of math correction, describing how GPT-5.5 Instant “caught the error” in a user’s rearranged quadratic after GPT-5.3 initially agreed with the user’s math.

Alongside accuracy, OpenAI is emphasizing personalization controls, including a new “memory sources” feature that is meant to show users what context was used in a response.

“Skip to main content 6 hours ago - Technology OpenAI makes default ChatGPT more personal Megan Morrone email (opens in new window) sms (opens in new window) facebook (opens in new window) twitter (opens in new window) linkedin (opens in new window) bluesky (opens in new window) Add Axios as your preferred source to see more of our stories on Google”

Axios

The New Stack says OpenAI is “introducing ‘memory sources’ across all ChatGPT models,” and it explains that these “show which inputs were used in a response and allow users to remove them.”

The New Stack adds a specific user-facing framing: “When a response is personalized, you can see what context was used, such as saved memories or past chats, and delete or correct it if something is outdated or no longer relevant,” quoting OpenAI’s blog post.

Axios similarly says OpenAI is adding “memory sources,” describing it as “a control that shows users some of the context ChatGPT used to personalize an answer, such as saved memories or past chats,” and it notes that OpenAI says users can “delete or correct outdated memory and use temporary chats that do not use or update memory.”

The Decoder expands on how the feature works, stating that “memory sources” show users which stored context shaped a reply, and it says “Entries can be flagged as relevant or irrelevant, edited, or deleted.”

Decrypt also quotes OpenAI’s guidance on control, saying “You remain in control of what's in your memory,” and it adds that “Temporary chats still opt out entirely.”

Axios warns that the increased memory “serves as a reminder that prompts may be stored by AI services, depending on the service and settings,” and it ties that risk to connecting third-party services like Gmail.

OpenAI’s rollout is described as both broad for the default model and staged for personalization features, with multiple outlets specifying which plans get access first and how long GPT-5.3 remains available.

The New Stack says GPT-5.5 Instant “is rolling as the default model in ChatGPT now,” and it also notes that GPT-5.5 Instant is designed for “everyday tasks such as writing, analysis, and general queries.”

CNBC

TestingCatalog says GPT-5.5 Instant begins rolling out “today to all ChatGPT users” and is also available in the API as “chat-latest,” while paid users keep GPT-5.3 Instant for three months through “model configuration settings.”

Mashable says GPT-5.5 Instant “has begun rolling out to all users as the new default model,” and it adds that GPT-5.5 Instant is “available to everyone,” while GPT-5.5 itself and Claude Opus 4.7 are described as only available to paid customers.

Axios specifies that enhanced personalization is initially for “Plus and Pro on the web,” and it says “Free, Go, Business and Enterprise will come later,” while also stating that paid users can keep GPT-5.3 Instant for “three months.”

TestingCatalog similarly says personalization from past chats, files, and connected Gmail is rolling out first to Plus and Pro users on the web, with “mobile support coming soon” and expansion planned for Free, Go, Business, and Enterprise users “in the coming weeks.”

The Decoder provides additional detail on the memory sources rollout, stating that “Memory sources will roll out to all consumer plans on the web first, with mobile to follow,” and it also says “Free, Go, Business, and Enterprise plans are expected to get access over the coming weeks.”

Several outlets describe what GPT-5.5 Instant is intended to feel like in day-to-day interactions, emphasizing shorter, tighter answers and reduced unnecessary conversational clutter.

The New Stack says the emphasis is “less on big bells-and-whistles features than on how ChatGPT behaves in day-to-day use,” and it frames the goal as making the model “be liked.”

CXOToday

Mashable says OpenAI’s blog post describes “stronger and tighter answers across subject areas,” “a more natural conversational tone,” and “better use of the context you’ve already shared when personalization can help.”

The Decoder adds that OpenAI focused on cutting fluff, saying the model “delivers the same information, often with more utility than previous models, while reducing the verbosity and overformatting that can make responses too long.”

Axios similarly says Instant avoids “unnecessary follow-up questions and formatting that made responses feel cluttered,” and it ties that to casual advice prompts.

Decrypt adds a specific example of how the model handles a math workflow, describing how GPT-5.3 initially agreed with a user’s solution but then “wrongly concluded there was no real solution,” while GPT-5.5 Instant “caught the error” and solved the corrected quadratic.

Axios adds a cautionary note that the increased memory “could increase people's reliance on it,” while also increasing “its access to their lives,” and it frames the privacy tradeoff as a continuing tension when connecting third-party services like Gmail.