preamble
After the Chief AI Sharing Circle recommended numerous free big model API services, an important issue suddenly came to light:
There is an official free-to-use small size scale; there is a reverse API model; but there has been no free "official conversion" API.
The reason I haven't recommended it is that free "official" large model APIs are difficult to provide stable and long-term services, and may even induce you to buy low-quality paid services.
Implications of providing an API for the "official transfer" model
There are some natural language tasks that can only be performed properly with a large-sized model of the task
Some AI apps allow only the "official to" large model API to be filled in when setting up a large model.
What is the "official transfer" large model API?
Officially provided APIs for big models from OpenAI, Azure, gemini or anthropic are packaged into their own API format (uniformly compatible with the OpenAI standard format) and sold through the oneapi/newapi relay platform. This is when the bright side is explained.
The real situation is: many "official forwarding" service providers use "black card", "reverse" or "tragic false" way to reduce cost. The real situation is: many "official forwarding" service providers use "black card", "reverse" or "tragic fake" to reduce the cost, so you can buy a large model API at the price of 1 RMB to 1 USD. of course, there are also some purely official forwarding service providers, which is very difficult for you to come across and judge.
The advantages of the "official" big model API are also obvious: 1. high load service, 2. low cost, 3. easy to buy domestically, 4. one-time access to the model comprehensively
Main article:Apply for a free "official conversion" API for the full range of OpenAI GPT models
access address
Registration Tips
Please use three-way login to register, to improve the success rate of registration
Creating a Token
Azure Public Benefit must be selected, and this channel is represented as an Azure forward only.Not an Azure-formatted APIThe
Validating API usable models
Access:Large Model Transit API True/False Validator , enter the interface address and key of the application number from the previous step.
First get the model list, then check the model and click "Test Model", wait for 30~120 seconds, you can check the validation result.
Non-real test results, the actual available models are very full, note that the actual use to avoid the use of gpt4-32K.
How to use
Since it is a standard OpenAI interface format, no further details are needed. Demonstrate the method of configuring it in Dify.
How Nextchat is configured
