4 hours ago
GPT-5 is here: Is OpenAI's new 'PhD-level' model up to the hype?

Web Desk
|
8 Aug 2025
OpenAI has launched the fifth generation of its artificial intelligence technology that powers ChatGPT.
The model, released on Thursday, arrives more than two years after the debut of its predecessor, GPT-4, in March 2023.
Its release comes at a time of heightened scrutiny and enthusiasm surrounding generative AI, amid both commercial investment and growing public debate about the risks and societal impact of the technology.
GPT-5 is here.
— OpenAI (@OpenAI) August 7, 2025
Rolling out to everyone starting today.https://t.co/rOcZ8J2btI pic.twitter.com/dk6zLTe04s
Speaking at a livestreamed event, OpenAI CEO Sam Altman said GPT-5 represents “a significant step along our path to AGI,” adding that the new model is designed to be both more powerful and more accessible to users.
“It’s like having a PhD-level expert at your fingertips, in any field, any topic, whenever you need it,” Altman told viewers, highlighting usability improvements for more than 700 million people who use ChatGPT weekly.
The release also comes just days after rival firm Anthropic unveiled an upgraded version of its Claude chatbot, underlining the fierce competition among tech companies in the race to lead the next wave of AI development.
OpenAI, headquartered in San Francisco, has evolved significantly since its founding as a nonprofit research lab in 2015. The company now operates under a for-profit model, currently valued at around $300 billion, while continuing to position itself as a mission-driven entity committed to safe AI development.
Its journey, however, has not been without controversy. In late 2023, Altman was briefly removed by OpenAI’s nonprofit board before being reinstated amid internal tensions over governance and direction.
The company is also facing regulatory scrutiny, with attorneys general in California and Delaware examining its transition from nonprofit status, and a legal challenge from Tesla CEO Elon Musk, who helped establish the organisation and has since accused it of straying from its founding mission.
Here is how netizens reacted to the launch of GPT-5:
so gpt-5 ...
— Lance Martin (@RLanceMartin) August 7, 2025
+ good for code (SOTA SWE-bench)
+ good for agents (SOTA Tau2-Bench)
+ worse than GPT-4o/-4.5 at writing
+ good pricing (below)
+ we have jobs through 2030 till further notice pic.twitter.com/rMACeIRNeC
Super disappointed in GPT-5. Poor prompt adherence, I have to bully it into giving me what I asked for.
— Riccardo Spagni (@fluffypony) August 8, 2025
It performed terribly on a simple image recognition task (extracting song names from screenshots) - it gave me 32 out of the 39 songs on the list. Claude Opus 4.1 nailed it…
Super disappointed in GPT-5. Poor prompt adherence, I have to bully it into giving me what I asked for.
— Riccardo Spagni (@fluffypony) August 8, 2025
It performed terribly on a simple image recognition task (extracting song names from screenshots) - it gave me 32 out of the 39 songs on the list. Claude Opus 4.1 nailed it…
Comments
0 comment