Do you want smarter insights into your inbox? Register for our weekly newsletters to get only what is important for Enterprise AI, data and security leaders. Subscribe now
OpenAi co-founder and CEO Sam Altman publicly recognizes large hiccups in the roll-out of GPT-5 yesterdayThe new, Large Language Model (LLM) flagship of the company – advertised as its most powerful and capable to date.
Answer user questions In a Reddit Ama (ask me everything) thread And In a message on x this afternoon, Altman has admitted to a number of problems that have disrupted the launch of GPT-5, including defective model circuit, poor performance and user confusion – Encourage OpenAi to partially reclaim part of its platform changes and Recovery user access to earlier models such as GPT-4O.
“It was a bit bumpy than we had hoped,” wrote Altman in Answer to a question on Reddit Regarding the large GPT-5 launch.
In terms of Incorrect model performance cards Shown during OpenAI’s GPT-5 Livestream, Altman said: “People worked late and were very tired, and the human mistake stood in the way. A lot comes together for a live stream in the last hours. “
Ai -scale distribution touches its limits
Power caps, rising token costs and inference inference reform Enterprise AI. Become a member of our exclusive salon to discover how top teams are:
- Change energy into a strategic advantage
- Architecting efficient conclusion for real transit profits
- Unlocking competitive ROI with sustainable AI systems
Secure your place to stay ahead: https://bit.ly/4mwgngo
Although he noticed that the corresponding blog post and system card were accurate, the missteps also strengthened a launch that has already been confronted with the control of early users and developers.
Problems with a new automatic model router
An important reason for the problems according to Altman stems from the new automatic “router” of OpenAI that assigns user prompts to one of the four GPT-5 variants-ReguliE, Mini, Nano and Pro-with an optional “Think” mode for heavier reasoning tasks.
On X” Altman revealed that an important part of that system was the car switcher “out of use for part of the day”, which meant GPT-5 “Way Dumber” then meant.
In response, OpenAI says that the changes to the model decision -making limit inquires and it will make it more transparent which model responds to a certain query.
An onion update is also on its way to help users activate the thinking mode manually.
In addition, Altman confirmed that With OpenAI, ChatGPT Plus users can now continue to use GPT-4O the earlier standard model -After a wave of complaints about the inconsistent performance of GPT-5. He said on Reddit The company is’Try to collect more information about the considerations ”before you decide how long offering legacy models.
Many users, including OpenAI BETA Testers such as Wharton School of Business Professor Ethan Mollick pronounced confused and horror on OpenAI an ilateral upgrading their chatgpt experiences to GPT-5 and initially removing access to the older models.
Real-World Performance are lagging behind Hype
The internal benchmarks from OpenAI can show GPT-5 that leads the LLMS packageBut real users share a different experience.
Since the launch, users have placed countless examples of GPT-5 that make basic errors in mathematics, logic and coding tasks.
Data Scientist Colin Fraser placed screenshots of GPT-5 wrongly resolving or repeating 8,888 is equal to 9 (that is of course not)While another user showed that it is flubbering a simple algebra problem: 5.9 = x + 5.11.
And other users reported problems to get accurate answers to mathematical word problems or the use of GPT-5 to debug his own presentation cards.
Developer Feedback has not been much better, with Users post Images of GPT advisory worse at “One-shot” certain programming tasks -Y well completed with a single prompt- Compared to rival AI Lab Anthropic’s new model Claude Opus 4.1.
And security company SPLX has found that GPT-5 still suffers from serious vulnerabilities To cause injection and to cover up the logical attacks, unless the safety layer is paved.
OpenAI in the spotlight
With 700 million weekly users on Chatgpt, OpenAI remains the largest player in generative AI per audience.
But that scale has brought growth pains. Altman noticed that in his X post on API traffic doubled more than 24 hours after the GPT-5 launch, which contributed to platform instability.
In response, OpenAI says it will double the limits for chatgpt plus users and continue to adjust the infrastructure while collecting feedback.
But the early missteps-remedied by confusing UX changes and errors in a controversial launching Opened a window for rivals to get land.
The pressure is for OpenAi to prove that GPT-5 is not only an incremental update, but a real step forward. Many users are not yet convinced based on the first rollout.