OpenAI announced the launch of GPT-4o, a new iteration of its GPT model that introduces groundbreaking multimodal capabilities. GPT-4o builds on the company’s popular GPT-4 generation by advancing its abilities across text, images, and voice.
This release of GPT-4o had been highly anticipated and speculated about. Last week, OpenAI teased upcoming product updates in a livestreamed event, without confirming specific details. Previous reports had suggested the company may unveil an AI search engine or new model called GPT-5.
According to CTO Mira Murati, GPT-4o delivers “GPT-4 level intelligence” while vastly improving performance across different data types. It can seamlessly understand and generate content using any combination of text, pictures, and audio. This makes interaction with AI assistants like ChatGPT much more natural by allowing them to perceive users through multiple sensory inputs.
GPT-4o will empower ChatGPT with new vision and voice awareness. Specifically, users will now be able to ask the assistant questions about photos or code screenshots and receive instant responses. Moreover, they can also interrupt ChatGPT during responses and have conversations that incorporate the emotion and context of spoken dialogue. Additionally, Murati sees applications expanding over time, like ChatGPT being able to analyze and discuss live sports through video.
Under the hood, GPT-4o brings major upgrades. As a natively multimodal model, it can integrate diverse data sources more efficiently for unified reasoning. The API is also twice as fast as previous Turbo versions at half the cost, with double the rate limits for developers. However, due to safety reviews, voice functionality will initially launch only for select partners through a controlled alpha.
Access to GPT-4o starts rolling out today across ChatGPT, remaining free for all users. Subscribers will benefit from boosted message limits through the enhanced capacities of this groundbreaking next-generation AI. With its new omnidirectional abilities, GPT-4o heralds a future where human-AI cooperation can occur through increasingly natural cross-modal engagement.
EN 
FR