ChatGPT is becoming faster and more conversational as OpenAI upgrades its flagship GPT-5 model to GPT-5.1.
OpenAI announced two additions to the GPT-5 family: GPT-5.1 Instant and GPT-5.1 Thinking. Both models are now available in ChatGPT.
GPT-5.1 Instant, essentially the default and most frequently used model, is now “warmer, smarter, and better at following your instructions,” according to OpenAI. Meanwhile, GPT-5.1 Thinking is an advanced reasoning model that replies faster on simple tasks and provides more sustained reasoning on complex ones.
“We heard clearly from users that good AI should not only be intelligent but also pleasant to talk to,” OpenAI wrote in a blog post. “GPT-5.1 significantly improves both intelligence and communication style.”
The company added that both models give users a way to “shape ChatGPT’s tone,” allowing people to tailor how the chat platform responds depending on the conversation.
Both models launched for ChatGPT Pro, Plus, Go, and Business users, as well as the free tier. Enterprise and Edu subscribers will receive seven days of early access to the models before GPT-5.1 becomes the default. OpenAI said the models are also available via the API, both with customizable reasoning capabilities.
OpenAI noted that it will soon update GPT-5 Pro to version 5.1.
Instant and Thinking models
The 5.1 tag reflects improvements to the base model, and OpenAI positions these models as part of the GPT-5 family, trained on the same stack and data as its reasoning models. The biggest difference between 5.1 and 5 is a more natural, conversational tone, said Fidji Simo, OpenAI’s head of applications, in a Substack post.
“In early tests, it often surprises people with its playfulness while remaining clear and useful,” OpenAI wrote.
Instant can use adaptive reasoning to decide when to spend extra effort on responses, particularly for more complicated queries. OpenAI said it improved the model’s instruction-following, so while Instant remains fast, it also directly answers the user’s question.
Recently released model versions, such as Baidu’s ERNIE-4.5-VL-28B-A3B-Thinking, have outperformed GPT-5 on instruction-following benchmarks.
GPT-5.1 Thinking can decide autonomously how much reasoning power to apply to a prompt. It adapts to the type and complexity of a question, taking longer on fuller, more complex requests than on simple summarization tasks.
OpenAI said evaluations showed that GPT-5.1 Thinking spends less time and therefore uses fewer tokens on simple tasks compared with GPT-5, outperforming the base model in response speed.
One point for organizations to note is that GPT-5.1 Thinking responds “with less jargon and fewer undefined terms.” OpenAI said reducing jargon makes Thinking more accessible when explaining technical concepts.
More personalization
Another major update to ChatGPT is increased personalization. Users can now switch between a friendlier and a more authoritative chat experience in their conversations.
ChatGPT already allows users to choose preset tone options, but the update expands those options “to better reflect the most common ways people use ChatGPT.”
The options include “standard,” “friendly” (previously “listener”), “efficient” (previously “robot”), “professional,” “sincere,” and “quirky.” Two other personalities, “cynical” and “nerdy,” remain unchanged.
“We believe many people will find GPT-5.1 better at combining IQ and EQ, but one standard clearly cannot suit everyone,” Simo said. “That’s why we’re also making it easier to customize ChatGPT with a range of presets. The model has the same capabilities whether you choose the standard or one of these options, but the style of its responses will differ—more formal or familiar, more playful or direct, with more or less jargon or slang. Of course, eight personalities still don’t cover the full spectrum of human diversity, but our research shows many prefer simple, guided controls to too many settings or open-ended choices.”
You can also adjust how much ChatGPT uses emojis. OpenAI offers detailed controls for responses and is experimenting with options to make models more concise, warmer, or easier to scan.
A measured rollout
OpenAI’s rollout of GPT-5 was far from perfect. While executives, including CEO Sam Altman, praised the new model’s capabilities, the decision to initially retire older, well-loved models in ChatGPT met with dissatisfaction. Many early users also found GPT-5 did not outperform older options in areas such as math, science, and writing.
This prompted Altman to walk back some comments about removing models and to attribute performance issues to GPT-5’s router. The router, which automatically routes queries to the most suitable models, will remain in place, and GPT-5.1 Auto will route queries to the model type best suited to answer them.
OpenAI emphasized that GPT-5 models Instant, Thinking, and Pro remain available in ChatGPT’s model dropdown, although paying subscribers have only three months to compare these older versions with the 5.1 update. The sunsetting period for GPT-5 will not affect models such as GPT-4o.
“Going forward, when we introduce new ChatGPT models, our plan is to give people ample time to evaluate what changed and share feedback, allowing us to continue innovating our frontier models while ensuring a smooth transition,” the company said. “Sunset periods will be communicated clearly and well in advance.”