OpenAI’s Silent ChatGPT 4o Update, What You Need to Know

January 29, 2025 — OpenAI has released update to its GPT-4o model within ChatGPT, focusing on expanding the model’s knowledge base and enhancing its ability to interpret visual inputs.

Extended Training Data Cutoff Yay! 

GPT-4o’s training data now includes information up to June 2024, an extension from the previous cutoff of November 2023. This update enables the model to provide more relevant and contextually accurate responses, particularly concerning recent cultural trends and research developments. The fresher dataset also enhances the model’s efficiency in framing web searches.

Improved Image Understanding

The updated GPT-4o demonstrates a deeper comprehension of visual inputs, excelling in interpreting spatial relationships, analyzing complex diagrams, and understanding charts and graphs. These improvements are evident in its performance on multimodal benchmarks like MMMU and MathVista. Users can expect richer insights and more accurate guidance in areas such as spatial planning, design layouts, and visually driven problem-solving.

Enhanced Performance in STEM Fields

GPT-4o has shown marked improvements in mathematics, science, and coding-related tasks, achieving higher scores on academic evaluations like GPQA and MATH. Its enhanced performance on the MMLU benchmark reflects its ability to tackle more complex problems across various domains.

Increased Emoji Usage

In a lighter update, GPT-4o now exhibits more enthusiastic emoji usage, especially when users incorporate emojis into their conversations. OpenAI encourages users to provide feedback on this feature.

These updates underscore OpenAI’s commitment to refining ChatGPT’s capabilities, ensuring users have access to a more knowledgeable and versatile AI assistant.

You can follow and read the updates yourself here