What Is GPT? The Model Family, ChatGPT and the API
GPT is an OpenAI model family. This article distinguishes it from the ChatGPT app and uses the official catalog to explain model selection for text, image understanding and tool use.
Meet this model
GPT is the name OpenAI uses for a series of generative models. They generate text from input, and particular versions can also understand images when that modality is supported. ChatGPT is an application people use directly; GPT models can also be integrated into software through the API. These are distinct things. Specify the model ID and required features when choosing a model. Tool or input support in one GPT version does not imply support across every earlier version.
How the model name differs from the application
Uploading a report image and asking a question in ChatGPT uses an application that may also provide file processing and search tools. When developers call GPT, their own software must prepare the input and handle the response. The experience in the app does not establish how a particular API model will behave in every environment.
ChatGPT and the OpenAI API platform have separate billing systems. A ChatGPT subscription does not establish an API usage allowance for your own software. This article reviews official feature and service documentation. It does not benchmark GPT versions on identical tasks or infer API performance from subscription benefits.
Choose a catalog model for a specific task
As of October 20, 2026, the official catalog positions GPT-6 Astra for complex reasoning and coding, GPT-6.1 Sol for balancing capability with usage expenditure, and GPT-6 Luna for focused, high-volume tasks. These are the provider’s stated roles, not HotAI.news benchmark rankings. They do not imply that every workload should use the same tier.
For example, to extract order numbers from support records, prepare samples with missing fields and multiple orders, require a fixed JSON structure, and check that values come from the source. For changes spanning several code files, compare patch correctness and test results instead. Speed on classification does not establish accuracy on complex engineering tasks.
Tool integration and model lifecycle limits
GPT function calling can produce structured tool requests, but the application executes custom functions. To query inventory, provide an inventory tool and return its result to the model. A generated intention to look something up does not establish a connection to a company database or confirm that inventory was actually checked.
The GPT API described here is a hosted service. Integration instructions do not grant a license to download model weights. Older models may also be deprecated or shut down, so check the official lifecycle notices and replacements when maintaining an existing project. A fixed model ID helps comparisons, while retained input samples let you detect factual or formatting regressions after migration.
Checked sources
These sources support the concepts and methods described here. Examples and test scope are labeled separately; check the version of the tools you use.