I've been trying out various "reasoning" models (o1, R1, Gemini Thinking etc) ag...

		simonw 8 months ago \| parent \| context \| favorite \| on: OpenAI's o1 Playing Codenames I've been trying out various "reasoning" models (o1, R1, Gemini Thinking etc) against the NYT Connections word puzzle - it's a really interesting test of them. So far o1 Pro has been the most consistently successful: https://www.nytimes.com/games/connections

Wonder if they use llms to write those puzzles