OpenAI Unveils New Model, Widens Access to ChatGPT Tools
OpenAI gifts free ChatGPT users with new tools and capabilities, including a speedy voice assistant, announced at Spring Update event
OpenAI has unveiled a new flagship AI model and is widening access to its AI tools previously reserved for premium users.
The Microsoft-backed company showcased the new offerings during its Spring Update event.
Rumors swirled before the event that OpenAI was teaming with Apple to offer its foundation models in iOS 18.
Instead, the company announced it’s democratizing ChatGPT access. Having recently enabled ChatGPT use without a signup, OpenAI has now gifted free users the ability to build their own GPTs, as well as access the GPT Store and access to ChatGPT’s vision and voice capabilities.
Powering those now-free capabilities is GPT-4o, OpenAI’s fastest and most powerful to date.
“This is the first time that we are really making a huge step forward when it comes to the ease of use,” said Mira Murti, OpenAI’s chief technology officer. “This is incredibly important because we're looking at the future of interaction between ourselves and the machines.”
Mira Murti, OpenAI’s chief technology officer
There’s more to come as Murti teased: “Soon, we'll be updating you on our progress towards the next big thing.”
GPT-4o
GPT-4o (‘o’ being short for Omni) is the new flagship OpenAI model, usurping GPT-4 and GPT-4 Turbo.
The model is available to free and paid ChatGPT users — making it the most powerful model the company has made available for its non-paying chatbot users to date.
The model, which will be rolled out over the next few weeks, is twice as fast as GPT-4 and improves on its capabilities across text, vision and audio.
It matches GPT-4 Turbo in terms of performance on text in English and code and improves on the previous model in non-English languages.
The model works across all modalities and can support more than 50 languages.
Also available to developers from OpenAI’s API, the GPT-4o model is five times cheaper to run and has five times higher rate limits compared to previous models.
Despite being available for free, paid ChatGPT users will have access to five times the capacity limit on GPT-4o compared to free users. ChatGPT Team and Enterprise users will have even higher limits.
“The special thing about GPT four is that it brings GPT four level intelligence to everyone,” Murti said.
OpenAI touts its new model as powering more natural human-computer interaction.
The model is capable of accepting a combination of modalities for inputs and outputs, including text, audio and images.
OpenAI
The new GPT-4o has access to OpenAI’s memory feature, enabling it to remember a user's information and preferences across chats.
From an entirely separate window, users can ask the chatbot about chart data or discuss photos on a web page.
OpenAI
Refreshed User Interface, Voice and Vision Capabilities
ChatGPT’s underlying AI model got a revamp and so did its user interface (UI).
With ChatGPT now available as a desktop app, OpenAI has given the application a complete makeover, opting for a simpler, sleeker design.
But the real reason behind the UI update was a focus around making things seamless.
“The refreshed UI makes it simpler to use, much more natural as well,” Murti said.
ChatGPT can now be windowed on a desktop, letting users interact with it while using another application.
With wider access to ChatGPT’s voice capabilities, users can now have the chatbot open in a separate window and use it to help with problems seamlessly
Showcasing the voice capabilities, OpenAI staff demoed the chatbot handling a series of code from a separate window.
The user simply copies the code (control + C) and triggers ChatGPT using their voice. The chatbot was able to read the code without the user having to copy and paste it.
OpenAI
GPT-4o, which is powering the voice offering, can respond to audio inputs in as little as 232 milliseconds, the company claims. Previously, ChatGPT’s voice mode had a latency response time between two and five seconds.
Staff spoke to the chatbot and were able to interrupt ChatGPT and ask for a change in response and the bot nearly instantly responded.
The bot has a more natural-sounding voice too. OpenAI has been investing in its audio offerings as part of its Voice Engine project, as it’s attempting to create more natural-sounding speech.
App users can point ChatGPT at an object, like a piece of text or a drawing and interact with the chatbot about the contents, like in the live demo where they asked about simple crudely drawn math problems.
OpenAI
Dr. Andrew Rogoyski from the University of Surrey's Institute for People-Centred AI described the Spring Update as “another ‘Sora moment’ for OpenAI.”
“They made existing voice assistants look utterly primitive,” Rogoyski said. “OpenAI's conversational GPT-4o is impressive. The responses are very fast, a key feature in establishing engagement with the user. The responses were also pretty flawless in terms of a natural voice, including very human-like nuances and additions. The responses to unstructured questions were particularly lifelike. The overall demo was so good I was left wondering whether they'd rigged the demo and used a real human voice actor. I'm looking forward to trying GPT-4o.”
Alon Yamin, Copyleaks’ co-founder and CEO said with this latest release, ChatGPT has become an AI assistant, like Alexa or Siri.
“GPT-4o enhances capabilities across text, image and audio with real-time responsiveness, highlighting how OpenAI is intentionally putting the focus on the interaction between technology and humans, which can offer important feedback for possible future offerings, like a search engine,” said Yamin.
“Nevertheless, no matter how advanced the technology gets, keeping some skepticism around it is crucial because while it might be more advanced, it still isn’t without potential flaws, such as misinformation. The need for tools to set guardrails is becoming more essential as this technology evolves and becomes easier for the public to interact and integrate it into their lives.”
Read more about:
ChatGPT / Generative AIAbout the Author
You May Also Like