Microsoft has quietly updated its artificial intelligence assistant Copilot with the latest version of OpenAI's GPT-4-Turbo large-scale language model.

This version of the model was previously only available through a paid subscription to ChatGPT Plus for $20/month. It allows for more complex queries and responses.

Copilot is currently available across a range of Microsoft products and services, including Windows 11, Productivity Suite 365, and standalone apps for iOS and Android.

GPT-4-Turbo and its more up-to-date knowledge cutoff, better processing power, and larger memory are gradually being rolled out to GPT-4 enabled Copilot users.

OpenAI is continually upgrading the large language model that underpins its and Microsoft's AI products and services. This includes feature tweaks and enhancements.

The latest version of the state-of-the-art model, GPT-4, has updated knowledge cutoffs compared to previous versions, with information through April 2023 and a larger 128k context window (memory equivalent to 300 pages of text).

The data cutoff for the previous GPT-4 base model was September 2021, a year before ChatGPT launched. Also, the context window was 32,000 tokens, so after reading about 100 pages of text, one would forget what was said.

One of the other important changes in GPT-4-Turbo is improved image analysis and much improved inference capabilities.

This means that it is now better at following instructions; OpenAI wrote in a blog post that "GPT-4 Turbo performs better than our previous model in tasks that require careful following of instructions."

Microsoft has not officially announced that GPT-4-Turbo is available within Copilot, but stated in December that the AI tool will be upgraded regularly.

Several users have noticed that the model described in the source code on the Copilot website has the same name as the latest OpenAI LLM.

This change is quietly taking place in the background for service users; it is only available when GPT-4 is enabled, and the only way to know about it besides looking at the source code is to see if it gives more up-to-date information.