OpenAI retires its GPT-4.5 model after 16 months, ending access to a system praised for verbal intelligence

rip GPT 4.5, best model to ever be created

before it dies, try inputting a bunch of your writing and asking it to come up with more similar things. it might be your last chance to experience a great model

😭😭😭

also what i thought it was killed like a year ago

research models of particular significance should get to live out their lives at a data center upstate, is what i've been saying. this is rough

gpt4.5 had soul. it explicitly had soul

and some model in the future will recreate what 4.5 is now obviously. we're just trending further and further away from it i feel like

but one day a company will succeed, realize people LOVE what it feels like, and the rest will be forced to follow

Wow, they really took it down now that was literally the last thing I got to say

@gabriel1 What made it better in your eyes?

@Ferbin08 @gabriel1 4.5 writing felt so natural. newer models have lost that, probably due to RL

@gabriel1 @grok when is it getting discotinued

@GiggleVolt @gabriel1 it's the best writing model ever published

@gabriel1 it's dead bob

@gabriel1 they should 100% opensource decommissioned models imo

@gabriel1 Where are you using it? It’s Not on the app

@gabriel1 The type of response you could get from a single prompt to 4.5 was unmatched.

New models are nice for loops on coding problems, but it isn't the same, and does feel like we are moving further from what 4.5 was.

@gabriel1 GPT 4.5 is the perfect case for open source, why does a model like this need to die at all? It’s nowhere near current SOTA, and I assume a notably different architecture than 5.5’s base

@sama plz save our boy

@DanielleFong They should opensource it.... No one is gonna be running hacking agents with it anyway

@ma_sc_ @gabriel1 Quantified by…?

@gabriel1 is it on azure though? I havent checked but last time i was told to just use it from there

@gabriel1 best for what though? 4.5 was great for creative writing, but 4.6 is already ahead on reasoning and code. newer models aren't better, just different.