GPT
OpenAI's models and the API most tooling assumes.
- made by
- OpenAI
- source
- Official GPT site
Magic Ship is one shop in Vancouver, BC, working remotely with clients worldwide. We are not a partner, reseller, or certified vendor of OpenAI - we just build with this.
What GPT is
GPT is OpenAI's model family, served through the OpenAI API. The platform also covers embeddings, transcription, image generation, and a batch endpoint at reduced cost. Its request format has been copied widely enough to act as a de facto standard that other servers implement.
How we use it
We use GPT models where they win on the client's eval set, and their embedding models are a frequent default for retrieval because they are cheap, fast, and well understood. Because so much tooling speaks the OpenAI wire format, an OpenAI-shaped client is often also the simplest way to talk to a self-hosted vLLM endpoint. Model ids are pinned, and an upgrade goes through the eval suite before it reaches production.
Where it is the wrong choice
More than one request interface exists and the ecosystem is split across them, so examples you find may not match the one your code is on. For clients whose contracts or residency rules exclude US-hosted inference, it is off the table before quality enters the conversation.
Service lines it turns up in
Related tools
More in Model providers
Other tools in the same service lines
Building something on GPT?
Send the problem rather than a job spec. You get an answer on scope, on fit, and on whetherGPT is even the right call for it.
Start a projectGPT and OpenAI are trademarks of their respective owners, used here to say what we work with.