OpenAI has announced that the flagship model, GPT-4o, will be available to all ChatGPT users for free, with paying users receiving faster access. This update boasts improvements in text, voice and vision capabilities, as well as enhanced performance.

The “o” in GPT-4o stands for “omni”, denoting its capacity for natural interaction with input in the form of text, audio, and image, producing corresponding outputs. In practice, this means the future will bring real-time voice conversations and video interactions with ChatGPT, with plans to launch a new Voice Mode alpha in the coming weeks.

Additionally, GPT-4o has been red teamed by over 70 external experts in various domains such as social psychology, bias and fairness, and misinformation to identify and address risks associated with the new features. OpenAI acknowledges the safety concerns surrounding artificial intelligence and has committed to continuous efforts to mitigate potential risks as they emerge.

While text and image inputs and outputs are currently available, audio inputs and outputs will be developed and released with safety measures in place, including preset voices and adherence to existing safety policies. Further details on addressing the full range of GPT-4o’s modalities will be shared in the forthcoming system card.