I really hope Apple is working hard on improving on-device models for their use case so they can get out of this.
When I first read the headline, I thought they’d licensed a customized Gemma 3n for an on-device model.
Apple nears $1B Google deal for custom Gemini model to power Siri
21–30 of 49 posts
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#22Mind blowing they couldn't get this to work. It's struck me lately that the models don't seem to matter anymore, they're all equally good. The UX and integration with regular phone features is what makes the tool shine and by now there should be plenty of open source models and know how to create their own. What is Google offering that Apple can't figure out on their own? Maybe people don't personal assitant AI enoug…
I don't know, Gemini 2.5 has been the only model that's been able to not consistently make fundamental mistakes with my project as I've been working with it over the last year. Claud 3.7, 4.0, and 4.5 are not nearly as good. I gave up on chatgpt a couple years ago so I have no idea how they perform. They were bad when I quit using it.
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#23Coincidentally, Google pays Apple over a billion dollars a year (est. at 1.5B) to be the default search in iOS. Could be re-titled. Google closes their trade deficit to half a billion dollars per year.
It’s actually much more than that, $20 billion per year
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#24Mind blowing they couldn't get this to work. It's struck me lately that the models don't seem to matter anymore, they're all equally good. The UX and integration with regular phone features is what makes the tool shine and by now there should be plenty of open source models and know how to create their own. What is Google offering that Apple can't figure out on their own? Maybe people don't personal assitant AI enoug…
I don't know, Gemini 2.5 has been the only model that's been able to not consistently make fundamental mistakes with my project as I've been working with it over the last year. Claud 3.7, 4.0, and 4.5 are not nearly as good. I gave up on chatgpt a couple years ago so I have no idea how they perform. They were bad when I quit using it.
I use all of them about equally, and I don't really want to argue the point, as I've had this conversation with friends, and it really feels like it is becoming more about brand affiliation and preference. At the end of the day, they're random text generators and asking the same question with different seeds gives different results, and they're all mostly good.
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#25This is remarkable if you consider how much it must wound Apple's pride to make this deal with their main rival in the smartphone software space, especially after all the fuss they made about "Apple Intelligence". It's a tacit admission that Google is just better at this kind of thing.
> wound Apple's pride do businesses really "think" in a personified manner as this? isnt it just what the accounting resolves to as the optimal path?
The C-levels leading the companies might, and the tech CEOs in question have been at the helm for a long enough while to build up some emotional feelings.
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#26This is remarkable if you consider how much it must wound Apple's pride to make this deal with their main rival in the smartphone software space, especially after all the fuss they made about "Apple Intelligence". It's a tacit admission that Google is just better at this kind of thing.
I’m not sure. It could be a way to save a ton of money. Look at the investments non-Apple tech companies are making on data centers & compute.
Maybe paying Google a billion a year is still a lot cheaper?
Apple famously tries to focus on only a few things.
Still, they will continue working on their own LLM and plug it in when ready.
Edit: compare to another comment about Wang-units of currency
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#27If true this is the deal of the century. Apple pay 1/14th of a Wang per year for a top tier model whereas Meta burn multiple Wangs a year in salary alone and get garbage.
For those equally confused: Meta bought 49% of Scale AI for 14.3 billion, purportedly to largely bring Alexandr Wang on board.
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#28This is remarkable if you consider how much it must wound Apple's pride to make this deal with their main rival in the smartphone software space, especially after all the fuss they made about "Apple Intelligence". It's a tacit admission that Google is just better at this kind of thing.
> that Google is just better at this kind of thing
That might be true but Siri sucks so bad it doesn't matter. It uses GPT but the quality is OSS models' level.
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#29Why Gemini? Just because of the closeness between the two companies already or is there a technical reason? I like Gemini the least because each search results in slightly different hard to reproduce identically results... I find I like LibreChat the best and then just connect it to all the other LLMs like ChatGTP, Claude, Anthropic, etc. from there.
Re: Apple nears $1B Google deal for custom Gemini model to power Siri
#30Mind blowing they couldn't get this to work. It's struck me lately that the models don't seem to matter anymore, they're all equally good. The UX and integration with regular phone features is what makes the tool shine and by now there should be plenty of open source models and know how to create their own. What is Google offering that Apple can't figure out on their own? Maybe people don't personal assitant AI enoug…
I don't know, Gemini 2.5 has been the only model that's been able to not consistently make fundamental mistakes with my project as I've been working with it over the last year. Claud 3.7, 4.0, and 4.5 are not nearly as good. I gave up on chatgpt a couple years ago so I have no idea how they perform. They were bad when I quit using it.
Do you find that Gemini results are slightly different when you ask the same question multiple times? I found it to have the least consistently reproducible results compared to others I was trying to use.