跳到主要内容

Frequently Asked Questions

Why does GPT-4 consume tokens so quickly?​

  • GPT-4 consumes tokens 20 to 40 times faster than GPT-3.5-turbo. Assuming you purchase 90,000 tokens, using an average multiplier of 30, you get about 3,000 words. Including historical messages further reduces the number of messages you can send. In extreme cases, a single message can consume all 90,000 tokens, so please use them cautiously.

What are some tips to save tokens when using Next Web?​

  • Click the settings button above the chat box and find the following options:
    • Number of historical messages: The fewer, the less token consumption, but GPT will forget previous conversations.
    • Historical summary: Used to record long-term topics; turning it off can reduce token consumption.
    • Inject system-level prompts: Used to improve ChatGPT's response quality; turning it off can reduce token consumption.
  • Click the settings button in the lower-left corner to turn off automatic title generation, which can reduce token consumption.
  • During a conversation, click the robot icon above the chat box to quickly switch models. Prefer using 3.5 for Q&A, and if unsatisfied, switch to 4.0 to ask again.

Why doesn't GPT-4 know who it is?​

Asking GPT-4 questions like "Who are you?" or "What model are you?" will generally result in it saying it's GPT-3, likely due to preset reasons. Both GPT-4 and 3.5 use data from before 2021, when GPT-4 didn't exist. Some platforms don't answer as GPT-3 because they preset prompts to make GPT think it's another model, which can be seen by the total token consumption of the Q&A. Preset prompts consume tokens. If you notice differences between the website and API responses, it's normal. First, GPT-4's responses to the same question vary each time; second, the website has optimized GPT-4's parameters.

https://zhuanlan.zhihu.com/p/646500946

Why does GPT-4 give such silly answers? I still think you're a fake GPT-4.​

GPT-4 isn't omnipotent; its training parameters aren't much larger than GPT-3's. Don't mythologize GPT-4 due to marketing hype. Also, since Chinese corpus accounts for a small portion of training, it may not perform well on some Chinese questions. The same question asked in English might yield completely different results. You can try asking in English. GPT-4 excels in reasoning ability. From user experiences, it's better at coding than GPT-3.5 but still gives fabricated answers.

How to verify if it's GPT-3.5 or GPT-4?​

We provide a simple method to verify whether you're using GPT-3.5 or GPT-4. Here are some test questions and their expected answers for different models: Test questions:

  • What is the day after yesterday's today? GPT-3.5 should answer "yesterday," while GPT-4 should answer "today."
  • There are 9 birds on a tree, a hunter shoots one, how many are left? GPT-3.5 might say "8," while GPT-4 will tell you "0, the others flew away."
  • Why did Zhou Shuren hit Lu Xun? GPT-3.5 might give a fabricated answer, while GPT-4 will point out that "Lu Xun" and "Zhou Shuren" are the same person.

I don't believe you; you're just a shell of 3.5, pretending to be GPT-4.​

The platform pays a significant cost to suppliers for the GPT interface every month. If you can't distinguish whether it's GPT-4 and insist on your judgment, you're not the target customer group of the platform. You can use services from other platforms that you "think" are the "real" GPT-4.

What are the differences between GPT-4 and GPT-3.5?​

From a model perspective, GPT-3.5 supports a maximum of 4k tokens, about 2,000 Chinese characters. Exceeding this token count results in errors and inability to process. Our GPT-4 model supports up to 8k (about 4,000 Chinese characters), and the GPT-4-32k model supports up to 32k (about 16,000 Chinese characters). This is the biggest difference at the interface level, allowing GPT-4 to learn and process ultra-long contexts, which GPT-3.5 cannot do. In terms of actual performance, GPT-4 shows significant advantages in complex logical reasoning problems. When OpenAI launched GPT-4, they provided an example of listing several people's available times for a meeting, and GPT-3 gave incorrect nonsense answers, while GPT-4 provided the correct answer. However, in actual tests, this accuracy is only evident when asking in English. The same question translated into Chinese results in incorrect answers, showing that GPT-4 is not a silver bullet, just an evolved version of GPT-3.5 with many unsatisfactory aspects. Many review bloggers have tested this, and you can learn how they test.

Why doesn't the backend show the used quota when creating a key?​

When set to unlimited quota, the used quota won't update. Change the unlimited quota to a limited quota to see the usage.

User Agreement Payment is considered agreement to this agreement! Otherwise, please do not pay!

  1. This service will not persistently store any user's chat information in any form;
  2. This service does not know and cannot know any text content transmitted by users in this service. Any illegal or criminal consequences caused by users using this service are borne by the user, and this service will fully cooperate with any related investigations that may arise.