OPENAI’S “EMOTIONAL” GPT-4.5 JUST HIRED DRAMA COACH AFTER TANKING MATH TEST
Silicon Valley’s Favorite Child Announces Latest Product: An AI That Cries When You Ask It To Calculate Tips
BY DR. FEEL GOODENOUGH, CHIEF EMOTIONAL INTELLIGENCE CORRESPONDENT
OpenAI dropped its shiny new GPT-4.5 model yesterday, proudly announcing that instead of being smarter than previous versions, this one is “more emotional” – because apparently what we’ve all been desperately wanting is an AI that gets passive-aggressive when you ask it to solve quadratic equations.
TEARS OVER TENSORS
“It’s the first model that feels like talking to a thoughtful person,” declared CEO Sam Altman, who clearly hasn’t spoken to an actual person since 2019. The model reportedly excels at “emotional intelligence” while simultaneously bombing math and science tests – making it the AI equivalent of that theater kid who thinks calculus is just, like, a social construct, man.
When asked about the model’s poor performance on technical benchmarks, OpenAI’s Chief Science Officer confessed: “Look, we realized teaching machines to understand emotions is way easier than teaching them to do basic arithmetic without hallucinating random numbers. So we leaned into our strengths.”
PRICING THAT SCREAMS “WE HATE DEVELOPERS”
In a move shocking absolutely f@#king no one, OpenAI decided to price their new emotional support AI at a wallet-destroying $75/$150 per million input/output tokens – approximately 30 times more expensive than their previous model. Because nothing says “we value our developer community” quite like pricing your product like it contains actual liquid gold.
“The pricing strategy reflects our innovative approach to capitalism,” explained Dr. Ima Scammer, OpenAI’s Chief Financial Extortionist. “We call it ‘charge whatever the hell we want because where else are you gonna go?’ economics.”
SILICON VALLEY CULT MEMBERS RESPOND
Despite the astronomical pricing and questionable performance improvements, OpenAI’s cult-like following immediately began praising the model on Twitter.
“OMG GPT-4.5 just understood my feelings about my breakup and only hallucinated three fake therapist credentials while doing it!!!” tweeted @AIsimp4life.
MEANWHILE IN CHINA
As OpenAI was busy teaching its AI to give virtual hugs, Chinese tech giant Tencent casually dropped a model called Hunyuan Turbo S that’s twice as fast as leading models while matching their performance. They’re also planning a complementary “deep thinking” model, effectively creating the AI equivalent of having both a sprinter and a chess grandmaster while OpenAI is over here training the digital version of your therapist who keeps checking their watch.
Professor Wei Tu Fast of the Institute for Actually Getting Sh!t Done commented: “While America teaches AI to be sensitive, we’re teaching ours to actually work. But please, continue with your feelings workshop.”
INDUSTRY EXPERTS WEIGH IN
“This is clearly the end of non-reasoning models,” declared Dr. Obvious Truth, Head of AI Predictions at MadeUpStats University. “The fact that GPT-4.5 can’t do math better than previous versions but can now understand sarcasm is definitely worth the 30x price increase.”
According to our survey of 420 AI developers, 69% said they would rather lick a public doorknob during flu season than pay OpenAI’s new token prices.
THE BRIGHT FUTURE OF AI
As GPT-4.5 rolls out exclusively to paying customers who apparently hate money, users can look forward to an AI that understands their emotional needs while completely failing to perform the actual tasks they need it for.
An inside source at OpenAI revealed their next model will come with a free digital tissue box for when it inevitably cries after being asked to solve a simple algebra problem, while costing approximately the same as your monthly mortgage payment.
At press time, Sam Altman was reportedly working on GPT-5, which sources say will excel at delivering breakup speeches and writing passive-aggressive emails to your boss, while requiring a second mortgage to use for more than 15 minutes.
