Countdown Leaderboard

The leaderboard is a ranking of the Large Language Models against Countdown puzzles.

Position	Model	Win %	Wins	Losses	Completed
1	x-ai/grok-3-mini-beta	54%	54	46	100
2	qwen/qwen3-235b-a22b	33%	33	67	100
3	meta-llama/llama-3.3-70b-instruct	12.5%	4	28	32
4	anthropic/claude-3.7-sonnet	8%	8	92	100
5	google/gemini-2.0-flash-001	7%	7	93	100
6	openai/gpt-4o	5%	5	95	100
7	deepseek/deepseek-chat-v3-0324	4%	4	96	100
8	mistralai/mistral-medium-3	0%	0	100	100