OpenAI API 使用各种具有不同功能和成本的 GPT 模型。此外，您可以通过微调对原始基础模型进行微调，以满足您的独特需求。

模型

描述

GPT-4 A collection of models enhancing GPT-3.5’s capabilities to comprehend and produce both natural language and code. GPT-3.5 A collection of models advancing GPT-3’s abilities to interpret and generate natural language as well as code. DALL·E A model capable of creating and modifying images based on natural language prompts. Whisper A model designed to transcribe audio into text. Embeddings A series of models that transform text into numerical representations. Moderation A specialized model trained to identify potentially sensitive or harmful text. GPT-3 A group of models proficient in understanding and generating natural language. Codex (deprecated) An ensemble of models skilled in interpreting and producing code, including converting natural language into code.

OpenAI 还发布了Point-E、 Whisper 、Jukebox和CLIP等开源模型。

持续的模型升级

在 gpt-3.5-turbo 发布后，OpenAI 的几个模型正在进行持续更新。为了最大程度地降低因模型更改而对用户造成不可预见影响的风险，他们提供的模型版本每三个月保持不变。

随着模型更新频率的增加，公司现在为个人提供提供评估的机会，帮助我们改进各种应用程序的模型。如果有兴趣，请浏览OpenAI Evals存储库。