switch model
The API call permission for GPT 4 that I applied for two days ago has finally been applied for.
In the past two days, I have also stepped up development, so that the domestic website (aichatroom.cn) I built can quickly support the use of GPT 4.
The difference between GPT 3.5 and GPT4
GPT-3.5 and GPT-4 represent two different versions of natural language processing (NLP) models released by OpenAI. The main differences are as follows:
Model size: GPT-4 will have a larger model size than GPT-3.5. This means that the GPT-4 model has more parameters, which helps capture more patterns and structures in the training data. In general, models with more parameters are more expressive and perform better.
Amount of training data: Compared with GPT-3.5, GPT-4 has a larger amount of training data. This means that GPT-4 can handle more instances and scenarios, resulting in better generalization ability in various NLP tasks.
Performance: Due to the increased model size and amount of training data, GPT-4 generally outperforms GPT-3.5 on various NLP tasks. GPT-4 may be better at understanding and generating complex, long text, and problems that require more background knowledge.
Quality of generated results: The text generated by GPT-4 generally outperforms GPT-3.5 in terms of consistency, accuracy, and fluency. This means that GPT-4 is likely to generate more reliable and useful answers in more situations.
results of testing
Now many platforms in China claim to be able to use GPT4, but in fact there are very few developers who can apply for GPT 4 API call permissions, "many of them are pretending to be GPT 3.5 . " Use the following questions to distinguish between GPT 3.5 and GPT 4. You can see that the answers of GPT 4 are more "intelligent" and more accurate. GPT 4 will try to think from more angles.
GPT 3.5
GPT4
It can be seen that for the first and second questions, GPT 3.5 and GPT 4 have similar answers, but for the latter two questions, GPT 4 performed significantly better.
Advantages and disadvantages and applicable scenarios
Advantages and applicable scenarios of GPT3.5
The biggest advantage of GPT 3.5 over GPT4 is "cheap" . In some scenarios that do not pursue accuracy and do not require complex reasoning, using GPT 3.5 may be a better choice, such as: copywriting, mind map sorting, chatting, and knowing whether the answer is correct or not at a glance (for example, let me write a bash command or create sql).
Besides that, another advantage of GPT 3.5 is that it is potentially faster. The official website has optimized the speed of GPT 3.5 for plus users, which is faster than GPT 4.
Advantages and applicable scenarios of GPT4
The main advantage of GPT4 is "the answer is more accurate" , which greatly reduces the probability of serious nonsense like GPT 3.5. But compared to GPT 3.5, the API of GPT4 is much more expensive (30 times that of GPT 3.5). The plus membership on the official website is also as high as 20 dollars a month, and you can only ask 25 questions in 3 hours.
The applicable scenarios of GPT4 are questions that need to be answered more correctly, such as the need to understand more complex contexts, or in some relatively strong professional fields. At present, my own practice is to use 3.5 to ask again. If I can’t get a satisfactory answer, I will use GPT4 to ask again, and sometimes I will get the answer I want. But in reality, the difference in many problems is not too big.