OpenAI on Friday upgraded its artificial intelligence (AI) model GPT-4 Turbo, giving it new capabilities, especially in math, reasoning and writing abilities. The upgraded version of GPT-4 Turbo is now available to paying users of ChatGPT Plus, Team, Enterprise and API. The new AI model also comes with an updated database, with a knowledge deadline now announced as April 2024. Notably, the update comes just days after the artificial intelligence company announced new GPT-4 Turbo with Vision models in its API.

this announcement OpenAI’s official capabilities.” One area where users will see immediate improvements is in its conversational language. The company says responses will be more direct and less lengthy when writing using ChatGPT.

This was a complaint we encountered when comparing ChatGPT to Google’s Gemini. We found the latter to be more conversational, and generating content such as letters, emails, or messages felt more natural. In comparison, the responses from ChatGPT (which we tested on the publicly available GPT-3.5) felt overly formal and bland. It appears that a recent update has fixed this issue.

OpenAI also emphasized that the new model will provide better math, reasoning, and coding capabilities, but did not share any examples of improvements. Benchmark scores released by the company show significant improvements on the MATH and GPQA (Graduate Level Google Validated Questions and Answers) benchmarks. The HumanEval and MMLU (Massive Multi-Task Language Understanding) benchmarks corresponding to coding and natural language processing capabilities did not show any significant improvements.

Users will also see an updated knowledge base in the new GPT-4 Turbo model. The company has extended the data deadline to April 9, 2024, while the old Turbo model will only be updated until April 2023. Currently, the new AI model is being rolled out to all paying users of ChatGPT.

---