OpenAI Launches GPT-5.5 Instant as ChatGPT’s New Default Model
2026年5月6日
OpenAI has released a new foundation model called GPT-5.5 Instant, replacing GPT-5.3 Instant as the default ChatGPT model.
The update centers on reducing hallucinations (the generation of factually incorrect content) in sensitive areas such as law, medicine, and finance.
What’s Inside GPT-5.5 Instant
According to OpenAI, GPT-5.5 Instant maintains the low-latency performance of its predecessor while reducing hallucinations, with particular gains in law, medicine, and finance.
Benchmarks show meaningful jumps. On the AIME 2025 math competition test, the model scored 81.2 versus 65.4 for the previous model. On MMMU-Pro, a benchmark for image-and-text reasoning, it scored 76 versus 69.2.
The broader GPT-5.5 series was launched in April with improvements highlighted in coding and knowledge work. The new Instant version brings those gains to default-model users.
Stronger Context Management
A new headline feature is improved context management.
GPT-5.5 Instant can use its search tool to reference past conversations, files, and Gmail, and return more personalized answers. The feature is rolling out first to Plus and Pro users on the web, with mobile to follow. OpenAI plans to expand access to Free, Go Business, and Enterprise users in the coming weeks.
ChatGPT will also show users which memory sources contributed to each response. Users can delete outdated sources or correct mistakes. When a chat is shared, the memory sources are not visible to the recipient.
API Changes and the Older Model
For developers, the GPT-5.5 model is available through the API (application programming interface) as “chat-latest.” GPT-5.3 will remain as an option for paying users for three more months.
OpenAI has faced pushback when retiring older models in the past. After GPT-4o was deprecated, users pushed back strongly—some described the model as their “best friend” or “a mirror” and signed petitions to keep it. Even so, GPT-4o was officially retired in February 2026.
A similar reaction to GPT-5.3 Instant’s phase-out is possible, though the three-month extension for paying users provides a softer transition this time.
Implications for Enterprise ChatGPT Use
For enterprises that use ChatGPT in their workflow, the change raises several questions.
First, AI use in regulated areas (law, medicine, finance) becomes more practical when reduced hallucinations come as the default. Compliance teams may revisit the approval bar for AI usage.
Second, the cross-source context feature—pulling from past conversations, files, and Gmail—creates new questions about data governance. Decisions about which internal data AI may reference horizontally now sit on the agenda for IT and security teams.
Finally, the API design that pushes the latest model under “chat-latest” is convenient but introduces drift risk for production systems. Pinning a specific model version in production remains the safer pattern.
Key Takeaways
・OpenAI released GPT-5.5 Instant as ChatGPT’s new default, with a focus on reducing hallucinations in law, medicine, and finance.
・Benchmarks improved (AIME 2025: 65.4 → 81.2; MMMU-Pro: 69.2 → 76), and a new context feature references past conversations, files, and Gmail.
・Enterprises using ChatGPT should revisit compliance criteria, data governance for cross-source context, and version pinning in production.
Sources
TechCrunch (GPT-5.5 Instant release)
