OpenAI CEO Admits Mistakes in GPT-5 Launch
Sam Altman, the CEO of OpenAI, admitted the company made mistakes during the launch of GPT-5.
Users complained that GPT-5 felt less friendly, didn’t code as well, and replaced GPT-4o without warning.
To fix things, OpenAI brought back GPT-4o for paid users, increased GPT-5’s advanced reasoning limit from 200 to 3,000 prompts per week, and added updates to make GPT-5 more warm and friendly.
Altman said ChatGPT is now used by nearly 700 million people every week.
He also revealed that API usage doubled in two days, causing a shortage of graphics processors (GPUs).
OpenAI plans to spend trillions on new data centers and may launch new products, including:
- A possible bid for Google Chrome
- New apps led by Fidji Simo
- Investment in brain-tech company Merge Labs
India might become OpenAI’s biggest market as they work on cheaper access.
Developers say GPT-5 is good at technical reasoning but not as strong at coding compared to Claude Opus, Sonnet, and some Chinese models.
CodingPerformance #ModelComparison
Security experts also say GPT-5 has already been jailbroken, despite safety updates.
