OpenAI has introduced an updated version of its large language model GPT-4 Turbo to the paid versions of ChatGPT.
The company announced the update late Thursday post on X. “Our new GPT-4 Turbo is now available to paying ChatGPT users,” OpenAI explained. “We improved our writing, math, logical thinking and programming skills.”
GPT-4, a successor to the model that helped ChatGPT launch the latter service, debuted last March. Compared to its predecessor, the fourth generation LLM is better able to solve reasoning tasks and is less likely to output erroneous information. Six months ago, OpenAI introduced an updated version called GPT-4 Turbo, which allowed a further increase in response quality.
The company has since released several improvements to the model. The version rolled out this week for the ChatGPT paid tiers, gpt-4-turbo-2024-04-09, is the latest version.
The new model was trained on newer data than its predecessor. According to OpenAI, the previous version of GPT-4 Turbo only had access to information generated before April 2023. In contrast, the latest version has access to data created as recently as December.
OpenAI compared the two models using Simple Evals, an internally developed benchmarking tool that it has made available as an open source solution on GitHub. The tool assesses the accuracy of an LLM’s results by administering six sets of questions on dozens of topics. According to OpenAI, gpt-4-turbo-2024-04-09 performed better than its predecessor in all six question sets.
OpenAI observed the greatest improvement when it tested the new LLM on an assessment data set called GPQA. Developed by Google LLC, the file contains 448 graduate-level multiple-choice questions on topics such as math and physics. The latest version of GPT-4 Turbo answered 49.1% of questions accurately, an improvement of almost 10% over the previous version of the model.
The new LLM delivered a similar increase in accuracy in a separate benchmark test of math questions. For the four other question collections evaluated by OpenAI, the model showed improvements of a few percentage points.
According to OpenAI, gpt-4-turbo-2024-04-09 is accessible to users of ChatGPT’s paid Plus, Team, and Enterprise tiers. The first edition is aimed at consumers, while the latter two are aimed at businesses. ChatGPT Enterprise, the latest addition to the lineup, offers cybersecurity controls not available in the other versions as well as faster response times.
Reuters reported Today, OpenAI Chief Executive Officer Sam Altman recently hosted a series of events promoting ChatGPT Enterprise. According to the news agency’s sources, more than 300 executives from Fortune 500 companies attended the events. It is estimated that the consumer version of ChatGPT is already used by employees at more than 90% of Fortune 500 companies.