"All last week, OpenAI watchers reported seeing strange things. References to GPT-5.1 kept showing up in OpenAI's codebase, and a "cloaked" model codenamed Polaris Alpha and widely believed to have come from OpenAI randomly appeared in OpenRouter, a platform that AI nerds use to test new systems. Today, we learned what was going on. OpenAI announced the release of its brand new 5.1 model, an updated and revamped version of the GPT-5 model the company debuted in August."
"In their release notes for the new model, OpenAI emphasizes that GPT-5.1 is "smarter" and "more conversational" than previous versions. The company says that GPT-5.1 is "warmer by default" and "often surprises people with its playfulness while remaining clear and useful." While some people like talking with a chatbot as if it's their long-time friend, others find that cringey. OpenAI acknowledges this, saying that "Preferences on chat style vary-from person to person and even from conversation to conversation.""
OpenAI released GPT-5.1 as an updated, revamped version of GPT-5. GPT-5.1 is described as smarter and more conversational, with a warmer default tone and playful yet useful responses. Users can customize tone with presets such as Professional, Candid, Quirky, and Nerdy. The Nerdy tone may prompt more pedantic language. The update shifts toward consumer preferences, favoring conversational touches that everyday users enjoy. Enterprise users may prefer less playful behavior. GPT-5 previously delivered accurate, largely hallucination-free responses, and GPT-5.1 aims to further reduce hallucinations. References to GPT-5.1 appeared in OpenAI's codebase and a cloaked model codenamed Polaris Alpha briefly surfaced on OpenRouter.
Read at Fast Company
Unable to calculate read time
Collection
[
|
...
]