ErrataBench

ErrataBench, a benchmark created by revise.io, has tested 100 LLM variants across 3,196 runs to determine which models are the best proofreaders, with a total runtime of 9 days 3 hours 34 minutes and …