Maybe unpopular opinion but I think at this point SWE-Bench has done its part and we need a new benchmark because Gemini being on/near the same level as Claude is obviously wrong
I use both and think they’re comparable. AMA.
Gemini at the same level as Claude is believable. Gemini CLI is not at the same level as Claude Code.