Entraspan experiments with AI models from all the major companies

• Problems arise because there are so many models to choose from
• Often times free or near free is ideal for text generation
• Some models are prohibitively expensive and to be avoided
• The model lists may change, this list is cached for 24 hour intervals
• Lower case attribute names are from the api, mixed case is added

What a Mess!

• This is a work in progress, turned out to not be time consuming
• Prooves the business guys and technical side are not talking
• Claude is the worst maybe
  • No consistency between pricing and model id
  • Many models being returned with no pricing/info
• On chatgpt and claude I marked anything not matched as legacy
• Haven't attempted embeddings
• Would be nice to get this information directly from the api's

Augmented model attributes

• The model attributes often are missing, incomplete, or insufficient
• Would be nice if there were standard attributes to compare models
• That is what the purpose of this page is with a link to a json file
• The added attributes help applications/users select ideal models
• Moved chatgpt to the bottom cause of all the noise
• Moved xai to the top as a leading frontier level model