Will an open model with comparable results to GPT-4 on MMLU be released by the end of July 2024?
Mini
6
319
Jul 31
73%
chance

Must be direct prompting score; no chain of thought or reranking. Comparable means greater or equal to the GPT-4 overall MMLU score.

Open model means the weights are publicly available, including through a signup form. Note this is not the same as "open source".

Get Ṁ600 play money