I couldn’t wait to get my hands on the new Siri AI. If you read my iOS 27 review, you could probably tell it was one of my most anticipated features in the new iOS.
We all know Apple is like that friend who is always late to the party.
They’re usually not one of the first companies to jump on tech trends. Multiple-camera systems existed years before Apple adopted them. Wireless charging, high refresh rates, and widgets on the home screen are all things Apple added much later.
And AI followed the same pattern.
Apple eventually launched its AI efforts under the Apple Intelligence brand back in 2024, but the rollout came with plenty of challenges, from delays to features that didn’t arrive when originally promised. It sometimes felt like Apple just couldn’t figure out AI.
But recently, Apple finally rolled out a much-improved Siri with its new software update. Apple describes it as a completely reimagined Siri with personal context, onscreen awareness, web knowledge, and more systemwide app actions.
Sadly, I couldn’t test it initially because Siri AI isn’t supported on my iPhone 12 Pro Max.
Bummer, right?
But I didn’t take no for an answer. I quickly checked my laptop, and thankfully, my 2021 MacBook Pro with the M1 Pro chip was supported. I updated it, only to get hit with the “Join the waitlist” button.
I clicked it almost immediately.
It took more than three days before Siri AI finally became available on my laptop. (Pro tip to Apple: Is there any way to make this process faster?)
Anyway, these are some of the things I noticed while testing Siri AI.
Ogbonda Chivumnovu
Siri AI comes with a brand-new look and a dedicated app, and I love it
Right out of the gate, you can notice the changes Apple has made to Siri.
The new Siri logo basically screams that it’s a new Siri. It feels like Apple wanted to make a statement.
It’s bold, new, and innovative. I really like it.
The new app, though, looks and feels similar to chatbot apps like ChatGPT, Gemini, DeepSeek, Claude, and the rest. But, in true Apple fashion, it has the company’s minimalist design language, which can look a little uninteresting at times.
It has a history tab for your recent conversations on the left-hand side of the screen, along with a new chat icon at the top, much like most chatbot apps.
You can also upload documents for Siri to analyse, ask questions about them, and even summarise them.
Overall, I felt it was about time Siri had its own dedicated app. There are definitely things Apple could improve, such as giving users more control over different models and modes for text generation, image generation, and other AI tasks.
I obviously didn’t expect all of these features to arrive immediately, given that this is Siri’s first dedicated app. Still, it feels like Apple has a long way to go before Siri can properly compete with the big players in the AI space.
Siri AI is smarter than what came before, but the gap isn’t huge
The new Siri AI is powered by Apple’s next-generation Foundation Models, which run on-device and through Apple’s server-based Private Cloud Compute. Apple says its latest models were built in collaboration with Google and its Gemini models.
With that in mind, you would expect a noticeable improvement in intelligence.
And I did notice one.
Siri was able to handle some fairly complex tasks. For instance, I asked it to check my latest emails. It did that and even gave me a bullet-point list of them.
I then moved on to something a bit harder and asked it to reply to a particular email.
It did.
I also asked it to handle routine tasks such as messaging people and making calls, and it managed to do all of these without much hassle.
When it came to reasoning, Siri AI also did pretty well.
I asked it a couple of reasoning questions, and it answered most of them correctly, except for one.
I asked it to find the missing numbers in a sequence, and this is where things got interesting.
Siri provided the wrong answer initially. But when you looked through its response and workings, it actually gave the correct answer at the end.
Siri's answer to a reasoning question (image: Chivumnovu Ogbonda / Techloy.com)
I suspect the first answer was a case of the model jumping to a conclusion before properly working through the problem. Whatever the reason, it was a strange one because the correct answer was sitting right there in the same response.
Most of these questions were answered correctly by ChatGPT and Gemini without any similar hiccups.
On-screen awareness works, but it’s not revolutionary
Subscribe for free to continue reading this article
Subscribe SubscribeAlready have an account? Log in
