OpenAI has announced ChatGPT-4.5, the latest and largest version of its large language model, which it promises is its “most knowledgeable yet.”
The company says its latest LLM, which will be the last version of its ilk, before a switch to chain-of-thought reasoning, can provide more natural conversations than the breakthrough GPT-4 release a couple of years back.
The company says there are improvements across the board, with more flowing conversations and a better understanding of what the user is asking of it. The company also expects GPT-4.5 will make less errors (which it calls hallucinations) than its predecessor.
Get the Galaxy S25 while it’s cheap
Anyone who’s held out on upgrading to the Galaxy S25, your patience has been rewarded with this sizeable price cut courtesy of Voxi.
- Voxi
- Was £799
- Now just £720
View Deal
In a post on X, founder Sam Altman said the good news was: “It is the first model that feels like talking to a thoughtful person to me. I have had several moments where I’ve sat back in my chair and been astonished at getting actually good advice from an AI.”
The company says this model is “at the frontier of what is possible in unsupervised learning,” while the next major iteration will be more capable of human-level reasoning that would think like us, taking time to come up with a response rather than spitting out instant answers.
Altman added of v4.5: “This isn’t a reasoning model and won’t crush benchmarks. It’s a different kind of intelligence and there’s a magic to it iI haven’t felt before. Really excited for people to try it!”
ChatGPT-4.5 improvements
“By scaling unsupervised learning, GPT‑4.5 improves its ability to recognise patterns, draw connections, and generate creative insights without reasoning,” OpenAI wrote in the blog post announcing the update.
“Early testing shows that interacting with GPT‑4.5 feels more natural. Its broader knowledge base, improved ability to follow user intent, and greater “EQ” make it useful for tasks like improving writing, programming, and solving practical problems. We also expect it to hallucinate less.”
The so-called hallucination rate (with the AI getting something wrong) has now dropped to 37%, whereas it was 61% in GPT-4o.
The GPT-4.5 model does have the most up-to-date information from the web, but it doesn’t have “multimodal features like Voice Mode, video, and screen sharing” the company said.
How much does it cost?
OpenAI is offering ChatGPT-4.5 to its $200 a month Pro subscribers from Thursday February 27.
Altman said he’d wanted to roll this model out to Plus ($19.95 a month) and Pro users at the same time, but the company has literally run out of the GPUs needed to offer the technology to lots of people.
