Base Model Guide

Descriptions of Base models and their tradeoffs

Model	Description	Strengths
FLAN-T5	11B parameter T5 Model. Fine-tuned on instruction prompts.	Tasks with well defined language prompts. More lightweight than the biggest supported GPT models, while performing at the same quality.
GPT-Ada	Part of original, base GPT series. Ada is usually the fastest model.	Parsing text, simple classification, address correction, keywords.
GPT-Babbage	Part of original, base GPT series. Babbage can perform simple classification and is quite capable at Semantic Search ranking.	Moderate classification, semantic search classification.
GPT-Curie	Part of original, base GPT series. Curies is extremely powerful and very fast. Curie is quite capable for many nuanced tasks like sentiment classification and summarization. Curie is also good at answering questions and performing Q&A and as a general service chatbot.	Language translation, complex classification, text sentiment, summarization.
GPT-Davinci	Davinci is the most capable model family and can perform any task the other models can perform, with less instruction.	Complex intent, cause and effect, summarization for audience.
GPT-Turbo 3.5	Most capable GPT-3.5 model and optimized for chat at 1/10th the cost of text-davinci-003.	Optimized for chat but works well for traditional completion tasks.
GPT-4	More capable than any GPT-3.5 model, able to do more complex tasks, and optimized for chat.	Optimized for chat.
Cohere Generate XLarge	This endpoint generates realistic text conditioned on a given input.	Generally, smaller models are faster while larger models will perform better.
Cohere Generate Medium	This endpoint generates realistic text conditioned on a given input.	Generally, smaller models are faster while larger models will perform better.
AI21 J1 Jumbo	J1-Jumbo, with 178B parameters, is the largest and most sophisticated language model ever released for general use by developers.	Jumbo is the most capable model in the J1 family, but it's also the slowest and most expensive to run.
AI21 J1 Grande	High-quality, affordable language model at the convenient size of 17B parameters.	While Grande is significantly closer in size to J1-Large (7.5B parameters), a great majority of users have found that J1-Grande’s quality is comparable to that of J1-Jumbo (178B parameters). This is great news for all budget-conscious practitioners; J1-Grande, our mid-size model, offers access to supreme quality text generation at a more affordable rate.
AI21 J1 Large	J1-Large has with 7.5B parameters.	Smaller, faster and more affordable but overall less capable than Jumbo, though still very effective for many use-cases.