OpenAI’s Spring Update livestream on May 13 brought three major announcements from the AI company:

  • A new flagship AI model called GPT-4o.
  • A desktop ChatGPT app for macOS.
  • ChatGPT users who don’t pay for a subscription can now access more features for free.

The coming changes to ChatGPT “brings GPT-4 level intelligence to everyone, including our free users,” said OpenAI Chief Technology Officer Mira Murati during the livestream.

OpenAI CTO Mira Murati speaks in a livestream on May 13. Image: Screenshot by TechRepublic

GPT-4o improves on GPT-4 Turbo’s voice and video capabilities

Murati said OpenAI’s next flagship model GPT-4o is “faster” and “improves on its capabilities across text, vision and audio” compared to GPT-4. The “o” stands for “omni.”

Instead, GPT-4o responds faster than its predecessor GPT-4 Turbo and has increased proficiency in non-English languages, video and audio. GPT-4o will be 2x faster, 50% cheaper and have 5x higher rate limits than GPT-4 Turbo, Murati said. OpenAI CEO Sam Altman said today on X that GPT-4o is “natively multimodal,” meaning it can switch faster between voice, text and video analysis.

GPT-4o will be accessible for free within ChatGPT, and it will roll out over the next few weeks to users globally. ChatGPT Plus and Team users will be able to use GPT-4o first, while availability for Enterprise users is “coming soon,” Open AI said. Paid ChatGPT users will have up to five times capacity limits. GPT-4o will also be available in the ChatGPT API.

The demo of GPT-4o’s real-time voice capabilities in the ChatGPT app at the livestream sounded quite natural, including several examples of how the model will respond seamlessly when interrupted. ChatGPT sometimes struggled to distinguish what images it was supposed to be looking at, but its responsiveness was remarkable.

The company noted it wasn’t releasing GPT-5, the rumored next-generation model, during the Spring Update. Murati said there may be more news about upcoming releases “soon.”


Safety considerations around GPT-4o

OpenAI reassured people that GPT-4o has “new safety systems to provide guardrails on voice outputs,” plus extensive post-training and filtering of the training data to prevent ChatGPT from saying anything inappropriate or unsafe. GPT-4o was built in accordance with OpenAI’s internal Preparedness Framework and voluntary commitments to AI governance. More than 70 external security researchers red teamed GPT-4o before its release.

SEE: Organizations across the world adopt standards like Content Credentials to try to reduce misinformation and deepfakes created using AI. (TechRepublic) 

New ChatGPT desktop app lands on macOS

MacOS users will soon be able to download a ChatGPT desktop application. ChatGPT Plus users can access the desktop app today, while other free and paying users can expect to gain access to it “in the coming weeks,” OpenAI said.

This UI for the ChatGPT desktop application for macOS shows ChatGPT listening while a user works on code. Image: OpenAI

Refreshed UI

Concurrently with the new ChatGPT desktop application, ChatGPT apps and the desktop version will receive a new, cleaner UI intended to improve ease of use.

The refreshed UI for ChatGPT on a browser changes the position of the user settings menu and cleans up some iconography. Note the addition of a place to attach files within the prompt bar. Image: OpenAI

GPT Store and other capabilities now open to ChatGPT users for free

Starting May 13, ChatGPT users will see the following rolling out for people who don’t subscribe to paid plans:

  • Use GPT-4 instead of GPT-3.5.
  • GPT-4o can pull from both the ChatGPT model and the web.
  • ChatGPT can analyze data in charts and create new charts.
  • ChatGPT can chat about photos uploaded by the user.
  • Free tier users can upload text files that ChatGPT can then summarize, analyze or create new content from.
  • Free tier users can explore and use GPTs in the GPT Store.
  • Free tier users can take advantage of ChatGPT’s Memory feature, which lets the model remember previous conversations with the same user.

GPT-4o arrives amid heavy competition

Business users have no shortage of options for AI assistants today. Google quickly followed OpenAI’s demonstration of the ChatGPT voice assistant on May 13 with one of their own. On May 21, Google showed a video of the Gemini assistant with the ability to “see” and remember things in real time, potentially answering such questions as “Did I leave my glasses at the office?” Microsoft, meanwhile, is going all-in on AI PCs, bringing Copilot on-device. GPT-4o’s snappy speech will need to set itself apart through usefulness and accuracy if it wants to succeed.

This article was updated to include competitors to OpenAI.

Subscribe to the Innovation Insider Newsletter

Catch up on the latest tech innovations that are changing the world, including IoT, 5G, the latest about phones, security, smart cities, AI, robotics, and more. Delivered Tuesdays and Fridays

Subscribe to the Innovation Insider Newsletter

Catch up on the latest tech innovations that are changing the world, including IoT, 5G, the latest about phones, security, smart cities, AI, robotics, and more. Delivered Tuesdays and Fridays