OpenAI has updated the base model in ChatGPT, with GPT-5.5 Instant taking over from GPT-5.3 Instant. The company has begun rolling out the update to all users and noted in a blog post that its new model is smarter, clearer, more concise, and more accurate compared to its predecessor.
What’s new with GPT-5.5 Instant?
OpenAI says GPT-5.5 Instant is more ‘dependable’, with the model making its biggest gains in terms of accuracy in internal benchmarks. The company says Instant produced 52.5% fewer hallucinations across areas like medicine, law, and finance when compared to GPT-5.3 Instant.
The model is also said to have produced 37.3% fewer inaccurate responses in challenging conversations that users had previously flagged for factual errors.
The ChatGPT maker also said that its AI is now more capable at everyday tasks like answering STEM-related questions, analysing image and photo uploads, and deciding when to use web search.
OpenAI also shared various examples of how the new ChatGPT model compared to its predecessor. In one such example, GPT-5.5 Instant was able to identify an algebra mistake, revisit the equation, and eventually arrive at the correct answer, while GPT-5.3 Instant arrived at the conclusion that there was no real solution.
“With this update, the model’s responses are tighter and more to-the-point without losing substance, while keeping the warmth and personality that makes ChatGPT enjoyable to use,” OpenAI said in its blog post.
The company also said that its chatbot now reduces unnecessary follow-up questions and avoids cluttering responses with gratuitous emojis. Notably, the company had recently revealed that it had faced a ‘Goblin’ issue in responses by ChatGPT that had also crept into its GPT-5.5 model. It’s not clear if that issue also persists with GPT-5.5 Instant.
In another example shared by the company comparing communication advice by two ChatGPT models, it noted that GPT-5.5 Instant used 30.2% fewer words and 29.2% fewer lines than GPT-5.3 Instant, while using an ‘informal, practical, and workplace-safe’ tone.
ChatGPT gets better memory:
OpenAI also says that its new model is more effective at pulling context from past conversations, uploaded files, and connected Gmail accounts. The model is also said to intelligently decide when a response can be improved with more personalisation and is faster at searching past conversations to find context, meaning users will not have to repeat themselves as often, or that’s what OpenAI says.
The company is also adding a new ‘memory sources’ feature in ChatGPT across all its models. With the new feature, whenever ChatGPT personalises a response, it will clearly show exactly what past context or saved memory it referenced to get there. OpenAI says users can easily delete or correct the memory if something is outdated or no longer relevant.
Moreover, memory sources will not be shown to others if users decide to share a chat with friends.
When can you use GPT-5.5 Instant?
GPT-5.5 Instant has already begun rolling out to all users and will soon replace GPT-5.3 Instant. For Plus and Pro users, the company says GPT-5.3 will remain accessible in model configuration settings for the next three months before it is retired permanently.
About the Author
