cursor Bench31

7 models evaluated

#	Model	Provider	Type	Score
1	Claude Opus 4.7 (Adaptive)	Anthropic	Closed	64.8%
2	Composer 2.5	Cursor	Closed	63.2%
3	GPT-5.5	OpenAI	Closed	59.2%
4	Composer 2	Cursor	Closed	52.2%
5	Gemini 3.5 Flash	Google	Closed	49.8%
6	Kimi K2.6	Moonshot AI	Open	47.6%
7	Kimi K2.5	Moonshot AI	Open	31.9%