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