OpenAI was quietly preparing Foundry, a new platform designed to let large customers run its AI models on dedicated computing capacity.
The service was expected to provide greater control over model configuration, performance, versioning and scaling, along with dedicated support and uptime commitments.
Pricing was aimed at large enterprises: running a lightweight version of GPT-3.5 was reportedly expected to cost $78,000 for three months or $264,000 for one year.
Documentation also referenced a model with a 32,000-token context window, fueling speculation that it could be related to the upcoming GPT-4.
Note: “GPT-3.5 Turbo” appears to be referring to the ChatGPT Turbo model.
https://docs.google.com/document/d/1EgPqpMYZRCxP4YFIbfZY9OWm1TykPzds8z5yUM13vbM/edit