Loading video...
Video Failed to Load
gpt4o in ChatGPT vs Apple Intelligence (on device Foundation model) prompt: "what are the top 3 coffee shops in lower manhattan?" not saying one is better than the other but fun to test. reply with your prompts and ill run a side by side for you.
248,720 views • 1 year ago •via X (Twitter)
8 Comments

My hot take is that the Apple model isn’t really meant for the traditional chat interface. It’s more for the types of examples where you automatically add tags to a post or detect the users tone or summarize a workout etc.

you're correct - apple says this as well in the wwdc videos. the above was for demonstration purposes.

🌆 The city never sleeps. ☕ The coffee's always hot. 💬 Who will you meet tonight? #OpenEndedNarrative Try the new chat sandbox, Stranger Sips, set in an all-night coffee shop. Where you go from there is entirely up to you!

This comparison is a little silly. On the left we have a very large (~200B) model *with* web search, running in a datacenter, and on the right we have a 3B model running on a phone - A much better comparison would be gpt4.1-mini (w/o web search) vs AFM!

not meant to be serious! was just curious.

why was 4o thinking wtf

How to get Apple Intelligence Chat?

Q: What are the top 3 coffee shops in Milas, Mugla, Turkey? The only answer I can accept: Do you aware of what you are asking about? Do you have a heatsroke? Do you want me to call an ambulance? All other answers are invalid 😅 Let’s check any GPT is intelligent enough
