Overview

To meet customers' diverse needs across various business scenarios, GPTBots offers a variety of LLMs and specialized models to cater to different business contexts. These models possess distinct capabilities (recognition and generation of text/images/audio/files) and characteristics, with service pricing varying based on the capabilities of each model. The types of model services provided by GPTBots are as follows:

• LLMs: Supports multiple versions of LLM models from various vendors, such as OpenAI, Claude, Gemini, ChatGLM, llama, and other mainstream commercial/open-source models.
• Embedding Models: Supports multiple embedding models, such as OpenAI, BGE, BCE, and other mainstream commercial/open-source models. However, in the SaaS version, to provide better services, only one embedding model is supported, but multiple vector dimensions can coexist.
• Rerank Models: Supports multiple rerank models, such as Jina, BGE, BCE, and other mainstream commercial/open-source models.
• Audio Models: Supports multiple audio models, such as OpenAI, Minimax, and other mainstream commercial/open-source models.

Terminology Definition

To facilitate enterprise users in managing and utilizing model services, GPTBots abstracts the services from the perspective of model brand and model version and defines the following concepts:
Model Vendor: The name of the company developing the large model, such as OpenAI, Anthropic, MistralAI, Ali, ZhiPu, Tencent, etc. Model Name: The brand name of the large model, such as GPT, Claude, Gemini, ChatGLM, llama, etc. Model Version: Versions of the large model with different capability quadrants, such as GPT-4o, GPT-o1-mini, Gemini-1.5-pro, etc.
Deployment Method: Model API services provided by the original vendor are official deployments, while those provided by third-party computing platform vendors are open deployments.
API Protocol: The protocol specification of the computing service platform running the large model, such as OpenAI running GPT, Amazon Bedrock running Claude, together.ai running Llama, and the relatively universal OpenAI protocol standard.
API KEY: Refers to the authentication configuration when calling the model service API. Model service providers usually impose frequency limits and independent billing for different API KEYs.

Model Service Configuration

GPTBots provides two ways to configure model services: using Platform KEY to configure model services and using Self-Owned KEY to configure model services. Enterprises can choose the method that suits their needs.
When configuring model services, GPTBots supports configuring multiple model API protocols and multiple API KEYs simultaneously for a specified "Model Brand/Model Version." When calling model services, GPTBots automatically rotates through different API KEYs to achieve high availability and high concurrency of model services.

Using Platform KEY to Call Model Services

Enterprises do not need to register accounts and purchase services from various model service vendors. They can directly use GPTBots' Platform KEY to call model services. Enterprises can view detailed token data of model service calls in "GPTBots Platform - Organization - Usage" and consume corresponding credits on GPTBots. For detailed pricing, refer to Service Pricing.

Using Self-Owned KEY to Call Model Services

Enterprises can host their registered and activated Self-Owned KEYs from various model service vendors on the GPTBots platform, paying related fees directly to the model service providers. Enterprises can view token data consumed by model service calls in "GPTBots Platform - Organization - Usage" and consume a small amount of channel service usage credits on GPTBots. For detailed pricing, refer to Service Pricing.## Model Service List GPTBots will continuously add model services suitable for enterprise business scenarios. If the model service required by the enterprise is not on the list, please contact us, and we will evaluate and test the model further to provide support. The currently supported model services and their capabilities are listed as follows:

LLMs

OpenAI GPT Series Models: GPT is a large language model series developed by OpenAI, including GPT-3.5, GPT-4o, GPT-o1, and others.

Anthropic Claude Series Models: Claude is a large language model series developed by Anthropic, including versions such as Opus, Sonnet, and Haiku.

Google Gemini Series Models: Gemini is a large language model series developed by Google, including versions such as Pro and Flash.