OpenAI's tight half-hour is teasing what might be announced ahead of the unveiling of its new GPT-4o AI model, prompting company CEO Sam Altman to manage expectations, saying, “It's not a search engine,” but we'll let you know. “Some New Things We Think People Will Like” After the event, AI enthusiasts and the tech press were both surprised and disappointed.
The release of GPT-4o – not GPT-5, as some people expected – represents a solid but incremental upgrade to GPT-4.5 Turbo. “omnimodel” works on text, audio and images, and offers enhanced coding capabilities, emotive female voices and faster generation times. Reactions to the news put GPT-4o and OpenAI on the global Twitter trending list for more than a day.
The event drew immediate comparisons to Google's Gemini Ultra, which was released in February. While many AI enthusiasts are excited about OpenAI's new model, some AI experts are less impressed.
Imad Mostak, the former CEO of Stability AI, shared a meme that sums up the reaction to OpenAI in IQ Scores. Much of the clamor has been over the debate over the inevitability of Artificial General Intelligence (AGI), but the long tail at both ends seems preoccupied with “AI waifus”—the virtual friends of lonely nerds.
The irony was easy to see as Stability AI's Stable Diffusion model was the preferred tool for making such friends.
OpenAI is particularly proud of the lifelike qualities of the voice they gave to ChatGPT, which Altman even joked is similar to “her” imaginary friend in a sci-fi movie. Andrei Karpati, founder of OpenAI, made a similar connection.
“The LLM killer app is Scarlett Johansson. You all thought it was math or something,” he tweeted.
Elon Musk, the mastermind behind Grok of “fun mode” fame, was less subtle in sharing his review. GPT-4o wrote in response to a Twitter user who was aggravating the “post-truth era,” “The demo disappointed me.”
Critics noted the presentation's lack of a single infrastructure feature, which mostly showed what could be done with current AI tools.
“I'm the same, but 20% faster,” Hugging Face founder Julien Chaumond tweeted. Yan Lekun, Meta's chief AI researcher, did not comment on his feelings about the OpenAI incident, but did post a laughing emoji in response to Chowmond's post.
Andrew Ng, co-founder of Coursera, praised the increase in efficiency: “The agent that used to take 10 minutes to switch APIs now takes only 5 minutes, which is a huge improvement.”
Others praised the potential practical applications of such a powerful model, which could be immediately tested in other forms. Imran Chaudhry, developer of Human AP, demonstrated a game using the device's GPT-4o, which highlights the model's visual capabilities—though latency is still an issue.
YouTuber Arun Maini praised GPT-4o for its smooth handling of subtitles.
But some enthusiasts, like novelist Ewan Morrison, see signs of a slowdown in AI development.
“The LLMs are proof that the plateau has already arrived,” he tweeted, “Open AI is making smoke and mirrors to prevent capital flight from investors.” He said he was puzzled why OpenAI didn't offer GPT-5, which was said to be released this year.
Digging deeper, Bindu Reddy, CEO of Abacus AI, posted benchmark test results that showed the GPT-4o performed worse than the GPT-4.5 Turbo on long and complex coding tasks, likely due to its multimodality.
Tim Simmons from the YouTube channel Theoretical Media summed up the situation in the middle: “I'm both overwhelmed and overwhelmed. While the GPT-4o has impressive capabilities, it is derided for its lack of truly novel features, saying the jump from the GPT-4 is less significant than its predecessor, the GPT-3.5.
Despite the mixed reactions, almost all of them drew comparisons between GPT-4o and popular sidekicks such as “She” or “Jarvis” from the Marvel superhero universe. Breaking down its language capabilities summary beyond “real-time voice translation,” Karpathy highlighted the value of OpenAI releasing native multimodal LLM rather than bridging between separate models.
Tanishk Matthew Abraham, director of research at Stability AI, says the live demo doesn't fully explain what's special about GPT-4o.
“Live streaming isn't the only cool part about GPT-4o,” he wrote. “No one is reading the GPT-4o blog post, which highlights many other great features,” including image text generation, visual consistency, text-to-3D capabilities, and video summarization.
“As for the AGI question, we're not even close,” said Jim Phan, NVIDIA's senior research manager. But it could be a new, more powerful GPT-5: Using the reported codename for the ongoing project, AI Leaker Jimmy Apples has hinted that the GPT-4o “feels like Gobi, or essentially Gobi,” and that there will be more announcements. In the near future from OpenAI.
The overpowered or boring GPT-4o lineup has stirred the pot and the public's attention is hoping to see a strong response from a top competitor at today's Google I/O keynote.
Edited by Ryan Ozawa.
Stay on top of crypto news, get daily updates in your inbox.
