ChatGPT has multiple AI models that you can use. The models each have their own strengths and weaknesses. While GPT-4o is the most versatile model for general use, there are many situations where the o1 model is the best choice.
As of this writing, OpenAI has released o3-mini and o3-mini-high. These aren’t complete replacements for o1 yet, but offer faster speeds and better coding capabilities, respectively. However, access to these new models may be more limited than o1 for the time being.
1
Use o1 to Solve Mathematical Problems
The o1 model is by far the best current ChatGPT model for solving complex mathematical problems. In fact, according to OpenAI, when given a qualifying exam for the International Mathematics Olympiad, o1 correctly solved 83% of the problems. In comparison, GPT-4o was only able to correctly answer 13%.
The beauty of o1 is that it doesn’t need to have been trained on a type of question to be able to answer it. It can apply different strategies to a mathematical problem, and discard those that don’t help until it finds a method that works.
If you need help with a complex mathematical problem, then o1 is the model to use.
Related
What Is ChatGPT’s o1 Model and How Can You Use It?
Use o1 to solve complex problems using step-by-step reasoning.
2
Try o1 for Help With Coding
If you need help with a coding problem, o1 is the model to turn to. While it still has some issues with hallucinations or outdated methods, it can be generally helpful when trying to figure out coding problems. In fact, it may come up with some solutions you haven’t considered.
It’s not got to the point where you can just sit back, type out what you need, and have o1 generate reams of flawless code. If you’re stuck, however, o1 can be a great place to turn to figure out what the problem is and how to solve it.
3
Get o1 to Generate Code
If you don’t have any coding skills at all, you can get o1 to generate code for you, as long as you’re not asking for anything too complex. For example, you can ask it to create a simple free-running game in which you use the space bar to make your llama character jump over cacti. Not only will it generate the code for the game, but using the Canvas feature, you can even test out the code right within ChatGPT.
You can use it for less trivial things, too. For example, you could create a design of an app, and get o1 to create the code to turn it into a working mock-up of your design. Once again, it’s not perfect, but you can achieve some useful results.
Related
Beginner Coding in Python: Building the Simplest AI Chat Companion Possible
You’ve probably heard of ChatGPT, but what about building a simple chatbot companion of your own?
4
Ask o1 to Solve Cryptic Crossword Clues
The o1 model’s reasoning skills aren’t just suitable for mathematical or coding problems, either. I’m a keen solver of cryptic crosswords, and I’ve always been intrigued by how well AI chatbots would handle them. Most models I’ve tried have been pretty terrible, often failing to follow the simplest restrictions such as the length of the solution.
However, o1 can often solve many cryptic clues, using its reasoning skills to try out theories and discard them when they don’t work until it reaches what it thinks is the most likely solution. It doesn’t always get them right, but it has an impressive hit rate.
You can use o1 for all sorts of other logic puzzles, too, although I tried it out on today’s Connections puzzle, and it took 51 seconds to get the right answer, which was quicker than me.
5
Get Help With Excel Using o1
Excel is a really powerful tool, but most of us barely touch the surface of what it can do. The investment of time learning all of its secrets just doesn’t seem worthwhile for most casual users.
If you need to do something complex with Excel, however, then the o1 model can be really useful. Not only can you ask it for help in figuring out how to achieve a goal using Excel, but it can even write Excel macros for you if what you’re trying to do is too complex to complete using standard formulas.
One thing the o1 model can’t do currently, which the GPT-4o model can, is create an Excel file for you. GPT-4o can create files, add data and formulas to them, and then provide a link for you to download the file. Currently, the o1 model is not able to do so.
6
Ask o1 How Many Times The Letter R Appears in Strawberry
This might seem trivial, but it was the first thing I tried when the first o1-mini model was released by OpenAI. That’s because, at the time, this was a question that GPT-4 was frequently stumped by. It would usually state that the letter R appears two times in the word strawberry when it actually appears three times.
This is down to the way that models such as GPT-4o are trained. Rather than thinking through problems, they simply output what is most likely to be the most useful response based on all of their training data. It won’t actually count the letters; it will just tell you what it thinks you’re most likely to want to hear.
The o1 model works differently; it will think about the way to tackle the problem (such as listing all the letters and counting each instance of the letter R) and then work through its chosen method before it answers. It meant that o1 was able to correctly answer the question when GPT-4o couldn’t.
The standard models in ChatGPT have been updated and are now able to get the strawberry question right, but it’s still possible to find ways to get it to generate incorrect answers because it doesn’t check whether the solution is correct before it gives it.
That’s why o1 is far better to use if your problem requires reasoning skills to solve. It may not be right every time, but it will at least think about its answer before it gives it. The downside of this, of course, is that o1 takes far longer to answer, especially for very complex problems.
The o1 model may be slow and ponderous, but what it lacks in speed it more than makes up for in sheer reasoning power. The ability to think through a problem, try different approaches, and assess which works and which doesn’t make it far more suited to solving complex problems that require some level of reasoning.
