Anthropic native Messages API

Updated 24-1123:

The official built-in PDF input function has been supported. To use this function, you need to add pdfs-2024-09-25 in the request header to enable this Beta function:

anthropic-beta: pdfs-2024-09-25, prompt-caching-2024-07-31

Notice:

Regarding Cache cache billing, if you need to use a large amount of cache, then because there are many Anthropic API accounts on the backend of this website to support the usage of all site users, and the Cache between multiple organizational accounts is not shared, in this case you may encounter multiple cache writes but no cache reads are triggered. This is not a bug. For this problem, we are designing a solution to allocate requests that may trigger cache reads to the same Key after the cache is written. After the Feature is online, a notification will be posted in the update log and TG notification channel.

0815 Update: Prompt Cache support has been added to some Claude models, which may save a lot of costs for continuous conversations with very long texts. This feature support is only implemented on the Messages native interface and will take effect with the corresponding request header. For details, please refer to: https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching

Native Claude Messages API compatible interface, input and output are provided as is, API version: anthropic-version: 2023-06-01

24-0405 Update: The latest ToolCall is supported

Note: The request method is slightly different from the native API:

Authentication method: The APIKey can be the same as the OpenAI API, and can be placed in the request header in the format of Bearer Token, or it can be placed in the x-api-key request header.

API version control: anthropic-version header native API is mandatory to fill in, otherwise an error will be reported. Although there is no mandatory requirement here, it is still recommended to fill it in. If not filled in, the latest version will be used by default, such as: 2023-06-01

The supported models are as follows:

claude-3-opus C3 large cup multi-modal

claude-3-sonnet C3 medium cup multi-modal

claude-2

claude-instant-1

For detailed documentation about this API interface, you can directly refer to Anthropic official Messages document

To use this interface, you only need to replace the API domain name and API key with those of this site and then you can debug and use them directly

For example, change https://api.anthropic.com/v1/messages to any of the following:

https://apic.ohmygpt.com/v1/messages

https://aigptx.top/v1/messages
*...

**Also, please pay attention to whether the API key permissions have checked the new model. This is because for security reasons, newly supported models will not affect all old keys. You need to manually check the permissions of these two models for the corresponding APIKey in the key management interface and save them, or create a new API key. **

That is, the current selling price = official price * Anthropic multiple: https://www.anthropic.com/api#pricing
The current limiting policy of this API service is: 6RPM for ordinary users, 60RPM for VIP users, and 480RPM for PREMIUM users.

curl --location --request POST 'https://cn2us02.opapi.win/v1/messages' \ --header 'Authorization: Bearer <token>' \ --header 'Content-Type: application/json' \ --data-raw '{ "model": "claude-3-opus", "stream": false, "messages": [ { "role": "user", "content": "Say test" } ], "max_tokens": 4096 }'

{ "completion": " 你好!我是一个由Anthropic公司开发的AI助手。我叫Clara。我被训练来进行友好、有帮助的对话。很高兴认识你!我希望我们的对话能让你感到开心和放松。如果有什么我可以帮到你的,请一定告诉我。", "stop_reason": "stop_sequence", "model": "claude-2.0", "stop": "\n\nHuman:", "log_id": "4fbf887cd2fd3b375d7da3b1cb81231b9401ff023772c8caf6d35bcac7ec9dd5" }

Anthropic native Messages API

请求参数

返回响应