GooseAi

GooseAI is a managed natural language processing API for text generation. It exposes open models, including GPT Neo, GPT J, GPT NeoX and Fairseq at sizes from 125 million to 20 billion parameters, through an interface that follows the OpenAI API format, so an existing integration can switch by changing the API base address. Text completion is available, and question answering and classification are listed as coming soon. It is aimed at developers who want hosted inference on open models at lower cost than the larger providers. The site describes the service as a joint venture of CoreWeave and Anlatan. Pricing is usage based and paid from prepurchased credits through a self serve platform. Each request has a base price covering the first 25 generated tokens plus a per token fee beyond that, with a configurable maximum of 2,048 tokens per call. In US dollars the base prices are $0.000035 for the 125M model, $0.000110 for 1.3B, $0.000300 for 2.7B, $0.000450 for 6B, $0.001250 for 13B and $0.002650 for 20B, with additional token fees from $0.000001 to $0.000063 per token. Bulk credit discounts are available to enterprise customers through the sales team.