ChatGPT just got a surprise update, but OpenAI can’t explain how it’s better

A new update to GPT-4o is out now

When you purchase through links on our site, we may earn an affiliate commission.Here’s how it works.

PerhapsSam Altman’s strawberry picturewas a secret message after all, as roughly a week after theOpenAICEO posted it on social media the company has confirmed thatChatGPTgot a surprise update last week that it hadn’t told anyone about.

Making this known on social media, an official ChatGPT account revealed that “there’s a new GPT-4o model out in ChatGPT,” adding in a follow-up that this isn’t a new model like a GPT-5 upgrade, just an improvement to the existingGPT-4omodel.

there’s a new GPT-4o model out in ChatGPT since last week. hope you all are enjoying it and check it out if you haven’t! we think you’ll like it 😃August 12, 2024

In itsModel Release Noteswe’re given a vague description of what these upgrades include – namely “Bug fixes and performance improvements” based on “experiment results and qualitative feedback” – which isn’t much to go on.

OpenAI admits that there’s not a new capability or a specific upgrade it can point to, even calling itself out for not having a quantitative benchmark which it can use to explain these smaller AI behavior tweaks. So we just have to trust it, and look at anecdotal reports that GPT-4o now works better than it did before.

It’s also unclear if this improvement is the‘Strawberry’project we’ve been waiting for. This is a codename which leaked earlier this year for the next ChatGPT model update, and while this latest improvement would count as an update, and may have been what Sam Altman was teasing with that strawberry image, leakers were expecting something a little more substantial for Strawberry.

It does, however, raise the question of how you track AI improvements. You can discuss the amount of data an AI has access to, or the processing power of the chipsets powering AI on devices, but what the end user cares about is how useful the AI actually is, which is a difficult thing to quantify as it’s based on people’s perception, rather than something you could measure. Here’s hoping someone (OpenAI or otherwise) comes up with an effective benchmarking tool soon.

i love summer in the garden pic.twitter.com/Ter5Z5nFMcAugust 7, 2024

Regardless, if you want to actually use the new ChatGPT model we have good news, it has already replaced the old version of GPT-4o inside the tool. So simply go to ChatGPT and you’ll be able to talk to the new bot just like you did the last version of GPT-4o – though for free users it’ll come with a message limit.

Get the best Black Friday deals direct to your inbox, plus news, reviews, and more.

Sign up to be the first to know about unmissable Black Friday deals on top tech, plus get all your favorite TechRadar content.

Alternatively, you might want to check out all of theGemini AIupgradesGoogleannounced for its upcomingGoogle Pixel 9phones. Features likeGemini LiveandAdd Meseem very impressive already, and we expect the tools will only get better.

You might also like

Hamish is a Senior Staff Writer for TechRadar and you’ll see his name appearing on articles across nearly every topic on the site from smart home deals to speaker reviews to graphics card news and everything in between. He uses his broad range of knowledge to help explain the latest gadgets and if they’re a must-buy or a fad fueled by hype. Though his specialty is writing about everything going on in the world of virtual reality and augmented reality.

How to delete a character from Character AI

How to turn off Meta AI

Belkin’s Travel Bag for Vision Pro has pockets and is way cheaper than Apple’s own case