GPT-4o

OpenAI has announced the release of ChatGPT-4o on May 13, 2024. This provides a true multimodal AI capable of natively understanding text, image, video and audio. This is faster than the GPT-4 model. This model has massive improvements in reasoning, processing and natural language capabilities. Its multimodal capabilities enable it to understand the images and it is capable of providing additional information about the image. In the future one can have a real-time conversation with the GPT model, which is one step towards making AI accessible and beneficial to everyone.

Its rollout is happening in a phase-wise manner across the globe. ChatGPT’s language capabilities further enhanced, it now supports more than 50 languages. GPT-4o’s capabilities will be rolled out iteratively. Its text and image capabilities are starting to roll out from May 13, 2024, onwards in ChatGPT. It will be available in the free tier as well. Its audio and video capabilities will be launched in phase wise manner for the small group of trusted users.

Developers can access GPT-4o in the API as a text and vision model. GPT-4o is 2x faster, half the price, and has 5x higher rate limits compared to GPT-4 Turbo.

Following are the features rolled out with the GPT-4o model –

• It can analyze the data and plot charts & graphs.
• It can chat about photos; it can also provide additional information about those photos as well.
• With GPT-4o, users will experience GPT-4 level of intelligence.
• It enables you to upload files, it can summarize/analyze the information available in files. It can also compare and contrast two documents.
• It enables users to create custom versions of GPTs that combine instructions, extra knowledge, and any combination of skills. Anyone can easily build their own GPT—no coding skill is required.
• OpenAI is also launching the GPT Store to help you find useful and popular custom versions of GPT models.
• OpenAI is building the ability for GPT to memorize things you discuss with it. So that future chats can take reference to those memories.

GPT-4o incorporates safety into its architecture across modalities, using strategies such as filtering training data and refining the GPT-4o’s behaviour via post-training. GPT-4o does not score above medium risk for the following categories - cybersecurity, CBRN, persuasion, and model autonomy.

OpenAI is beginning to roll out GPT-4o to ChatGPT Plus and Team users, and it will be available soon for enterprise users. Plus users will have a message limit that is up to 5x greater than free users, and Team and Enterprise users will have even higher limits.