Microsoft’s Copilot AI chatbot was already the best way to use GPT-4 for free. Now, it’s got even better by adding the more advanced GPT-4 Turbo model to its engine room.
The news that GPT-4 Turbo is beginning to replace GPT-4 as the large language model, or LLM, that powers the free version of Copilot is a big deal for a few reasons.
Namely, it shows that Microsoft is continuing to add value to its Copilot offering by giving everyone access to the latest and greatest AI chatbot capabilities, whether they want to pony up a month subscription fee or not.
By way of background, GPT-4 Turbo was previously only available on Copilot Pro, which costs $20 a month, or the same as OpenAI’s own ChatGPT Plus tier.
Now, here are the other reasons why it’s important, how it compares to the old Copilot with GPT-4. and of course how to use GPT-4 Turbo on Copilot yourself.
That’s because, as we’ve said, GPT-4 Turbo is more advanced than GPT-4, meaning in real world usage it will give you better responses, faster.
For starters, its context window is 128k tokens, compared to just 32k with GPT-4. In practice, this means that an AI chatbot powered by GPT-4 Turbo is able to process more than 4x the amount of text.
While many casual users may never get close to using its full context window, it means you’re able to prompt it with way more context to improve your output results, if you want.
It also has a more recent knowledge cut-off date, which in plain English means when you ask it something, it’s able to draw on more current information to informing its response. For GPT-4 Turbo, it’s pulling info up to April 2023, while older versions are restricted to January 2022.
Plus, as we’ve said, it’s all free – provided you know where to access GPT-4 Turbo in Copilot, of course.
How GPT-4 Turbo in Copilot Performs When Tested
We also wanted to see how Copilot with GPT-4 Turbo performed compared to GPT-4. To do this, I asked it to answer the popular tongue-twister: “How much wood would a woodchuck chuck if a woodchuck could chuck wood?”
Turbo was clearly superior, as GPT-4 initially didn’t even attempt to logically answer the query, however whimsical it might be.
GPT-4 Turbo, on the other hand, went all-in with the logical side of things and cited an appropriate reference in trying to solve the iconic if nonsensical query.
