The $300 billion outfit claims GPT-5 is now top of the class in science, maths and coding. It is being handed out to punters through ChatGPT, although free-riders will find themselves throttled pretty quickly.
This launch comes more than two years after the last model, GPT-4, raised expectations and valuations. Investors and the rest of the Silicon Valley faithful have been holding their breath waiting for this one.
OpenAI claims this version’s standout trick is something called “vibe coding” where users throw prompts at it and it spits out software like it is going out of fashion. “This idea of software on demand will be a defining part of the GPT-5 era,” Altman said.
ChatGPT launched nearly three years ago and dragged generative AI into the mainstream along with OpenAI’s spiralling valuation. The company now claims more than 700 million weekly users and, according to those familiar with the matter, is pitching for a valuation of $500 billion, which would crown it the world’s most overhyped private tech company.
Its backers have been banking on GPT-5 to justify that absurd number and prop up dreams of trillion-dollar paydays. Whether it is enough to appease the cocaine nose jobs of Wall Street remains to be seen.
Gartner analyst Chirag Dekate said: “GPT-5 meets expectations in technical performance, exceeds in task reasoning and coding, and underwhelms in [other areas].” He added that it “does not appear to cross the AGI threshold . . . and early observations seem to indicate incremental progress in writing quality.”
Coding continues to be one of the more useful things AI can do and GPT-5 just managed to edge out rival Anthropic in the SWE-bench Verified test, which is how the AI crowd measures coding prowess.
That result was enough to poach a few notable users. Michael Truell, chief executive at $10 billion outfit Anysphere, said GPT-5 was “remarkably intelligent.” His company makes Cursor, a coding assistant that normally leans on Anthropic’s models. If users jump ship, OpenAI’s already ridiculous annual recurring revenue, currently $12 billion, could balloon to $20 billion by the end of 2025.
GPT-5 has not exactly won every test though. Elon Musk’s xAI claim their Grok 4 Heavy still beats GPT-5 on certain reasoning and knowledge benchmarks. Although, Musk also said we would have affordable robots by the end of last year.
The new model appears less likely to hallucinate or clam up entirely, which is progress of sorts. It is also OpenAI's first flagship model to be slapped with a “high” risk label for potential bioweapon misuse.
The company said: “While we do not have definitive evidence that this model could meaningfully help novices to create severe biological harm, our defined threshold for high capability, we have chosen to take a precautionary approach.”
OpenAI makes its money charging individuals between $20 and $200 per month, with businesses getting stitched up on custom rates.
In what looks like a panic move to fend off competition, OpenAI also shoved out its first so-called “open” model since ChatGPT’s debut.
OpenAI ChatGPT product boss Nick Turley. ““The vibes of this model are really good. But at the end of the day, I think people are just going to have to feel it.”
So far, the product has not rolled it out to Italy so we have had a chance to play with it.
