▲ | swyx a day ago | |
analysis as the resident summaries guy: - sonnet has better summary formatting "(72.5% for Opus)" vs "Claude Opus 4 achieves "72.5%" on SWE-bench". especially Uncommon Perspectives section - sonnet is a lot more cynical - opus at least included a good performance and capabilities and pricing recap, sonnet reported rapid release fatigue - overall opus produced marginally better summaries but probably not worth the price diff i'll run this thru the ainews summary harness later if thats interesting to folks for comparison |