I’d expected Claude to be a clearer winner when we started this, but it was much closer, and the difference largely subjective. Both are exceptional when they have pre-calculated metrics.
Great article. I'd like to see this extended to the APAC and open models such as GLM and Minimax next time!