ChatGPT's updated default model is 50% less likely to provide misleading information on medical and financial inquiries.
GPT-5.5 Instant has replaced GPT-5.3 Instant as ChatGPT's default model, with the update being rolled out to all users today. The new model aims to enhance three key areas: accuracy, brevity, and personalization.
Fewer hallucinations, more brevity
According to OpenAI, the accuracy improvements are especially significant in critical fields. During internal testing, GPT-5.5 Instant generated 52.5% fewer hallucinated statements compared to its predecessor in prompts related to medicine, law, and finance, and produced 37.3% fewer inaccuracies on previously flagged conversations.
GPT-5.5 Instant is now being introduced in ChatGPT, offering smarter, clearer, and more personalized responses in a more engaging and natural manner. It also features shorter replies, which users had requested. We believe you’ll enjoy interacting with it. pic.twitter.com/HSQOhjqxp7— OpenAI (@OpenAI) May 5, 2026
Additionally, the new model improves its ability to analyze photos, address STEM inquiries, and make web search decisions compared to earlier versions. OpenAI claims that responses are now 30.2% shorter in terms of word count and 29.2% shorter in terms of line count than GPT-5.3 Instant.
OpenAI
Users can expect fewer unnecessary follow-up inquiries and improved formatting.
Enhanced memory, greater control
GPT-5.5 Instant also excels at drawing relevant context from past conversations, uploaded documents, and connected Gmail accounts to tailor responses without making users repeat information.
OpenAI
In conjunction, OpenAI is rolling out “memory sources” across all ChatGPT models. This feature will allow users to see what context was utilized to generate a response, with the option to delete or amend any of it directly in the settings.
The rollout of GPT-5.5 Instant will occur over the next two days for all ChatGPT users. GPT-5.3 Instant will remain accessible to paid users for three months via model settings before being phased out. The improved personalization features will first be available to Plus and Pro users on the web, with Free, Go, Business, and Enterprise users to follow in the coming weeks.
Memory sources will be introduced across all ChatGPT consumer plans on the web, with enhancements in personalization and memory sources set to come to mobile devices soon.
Pranob is an experienced tech journalist with more than eight years of experience reporting on consumer technology. His work has been…
Your ChatGPT history is an unintentional personality assessment
Researchers trained an AI to deduce your personality from your ChatGPT history, and it worked impressively well.
Whenever you request ChatGPT’s assistance in drafting an email, discussing a relationship issue, or searching for symptoms, you might be providing more than just a query. As reported by TechXplore, ETH Zurich researchers have developed an AI model capable of predicting personality traits directly from actual ChatGPT conversation logs, and it demonstrated a surprising ability to identify these traits accurately. The study gathered 62,090 real conversations from 668 ChatGPT users, and participants completed a standard personality test, giving researchers a reference for comparison. The AI was trained to categorize each user as low, medium, or high across the five personality traits: openness, conscientiousness, extraversion, agreeableness, and neuroticism.
iOS 27 may allow users to choose their preferred AI model for text and image tasks
Apple is not rushing to create the smartest AI; rather, it aims to provide the best platform for others to excel, and iOS 27 might just be the debut for a wide range of options.
A significant change in how AI functions across Apple devices is on the horizon, placing users at its center. According to Bloomberg’s Mark Gurman, iOS 27 and its major variations (iPadOS 27 and macOS 27) will enable users to select from various competing third-party AI models for core Apple Intelligence functions. Think of it as using your TV remote to set the desired channel—here, the remote is iOS 27, and the channels are some of the popular AI models you may have used elsewhere.
32GB RAM for Windows 11? Hey Microsoft, this seems like your issue!
If your operating system requires more memory, perhaps the issue lies not with the user's PC.
There was a time when buying a PC felt logical. Eight gigabytes of RAM sufficed, 16GB felt like a significant upgrade, and anything beyond that was for those engaged in truly heavy tasks. This balance existed because software was designed to acknowledge hardware limitations. Nowadays, that balance has discreetly shifted, and Microsoft appears to accept this change without question. Their previously retracted guidance, casually labeling 16GB as the new baseline and 32GB as the safe zone, represented more than just a suggestion—it was a shift in responsibility. Modern hardware does not demonstrate a need for double the memory
Other articles
AI-native enterprise expenditures jump by 94%, while SaaS growth remains flat at 8%, prompting a reevaluation of per-seat software pricing amid the SaaSpocalypse.
Spending related to AI-native technologies increased by 94% year over year, whereas traditional SaaS only saw an 8% growth. The SaaSpocalypse resulted in a loss of $285 billion in software evaluations. Each enterprise vendor is shifting towards agents.
Expenditure in AI-native enterprises has jumped by 94%, while spending on SaaS has remained stagnant at 8%, leading to a recalibration of prices for per-seat software in what is being termed the SaaSpocalypse.
AI-native spending increased by 94% year-over-year, whereas traditional SaaS growth reached 8%. The SaaSpocalypse resulted in a loss of $285 billion in software valuations. Every enterprise vendor is shifting towards agents.
Intel has showcased its Computex 2026 lineup, which includes handheld devices, desktops, and servers, as the 18A process emerges as a signature technology for the foundry.
At Computex 2026, Intel will present Panther Lake Arc G3 handhelds, a 52-core Nova Lake desktop, and a 288-core Clearwater Forest Xeon, all manufactured using the 18A process node.
The updated default model of ChatGPT is now 50% less likely to provide misleading information regarding medical and financial inquiries.
OpenAI has updated the default model for ChatGPT from GPT-5.3 Instant to GPT-5.5 Instant, which claims to reduce hallucinated responses on medical, legal, and financial inquiries by 52.5%.
The lineup for Xbox Game Pass in May 2026 has arrived, and it’s substantial, featuring Forza Horizon 6 and DOOM at the forefront.
Xbox Game Pass has announced 13 new games for May 2026, including major day one releases such as Forza Horizon 6, Mixtape, and Subnautica 2.
The surge of humanoid robots in China is encountering a reality check, as 150 companies pursue a market where merely 23% of customers express satisfaction.
China is home to over 150 humanoid robot companies, holding a 90% share of the global market; however, buyer satisfaction stands at just 23%. Morgan Stanley cautions that a shake-out may occur as the number of IPOs surpasses demand.
ChatGPT's updated default model is 50% less likely to provide misleading information on medical and financial inquiries.
OpenAI has substituted GPT-5.3 Instant with GPT-5.5 Instant as the default model for ChatGPT, assuring a 52.5% reduction in hallucinated responses regarding medical, legal, and financial inquiries.