Instructions following x Wrong answer Ranking

See which AI models are most likely to hit Wrong answer on Instructions following, so you can spot weak points faster. Sort by: Total Cost ↑.

Models Shown

Total Failures

111

Most Affected Model

Failure Reasons

Wrong answer111 Did not follow instructions73 API error25 Extra formatting8 Timed out5 No answer2

Categories

Domain specific878 Anti-AI Tricks502 Coding477 Puzzle Solving344 Trivia306 General Intelligence129 Agentic126 Combined126 Instructions following111 Data parsing and extraction93

111/111

Rank	Model	Company	Wrong answer Count	Category Score	Total Cost	Tests Correct	Response Time (avg)
#162	LongCat 2.0 high	Meituan	1	0.7	$0.778	1/2	6.96s
Total Tests 2 Wrong Tests 1 Total Cost $0.778 Response Time (avg) 6.96s
#292	Trinity Large Thinking high	Arcee AI	1	0.6	$0.794	1/2	4.12s
Total Tests 2 Wrong Tests 1 Total Cost $0.794 Response Time (avg) 4.12s
#267	Trinity Large Thinking medium	Arcee AI	1	0.4	$0.862	0/2	3.99s
Total Tests 2 Wrong Tests 2 Total Cost $0.862 Response Time (avg) 3.99s
#256	Trinity Large Thinking low	Arcee AI	1	0.4	$0.918	0/2	4.07s
Total Tests 2 Wrong Tests 2 Total Cost $0.918 Response Time (avg) 4.07s
#214	GPT-5.5 none	OpenAI	1	0.6	$1.212	1/2	1.15s
Total Tests 2 Wrong Tests 1 Total Cost $1.212 Response Time (avg) 1.15s
#144	GLM 5.1 medium	Z.ai	1	0.6	$1.546	1/2	7.47s
Total Tests 2 Wrong Tests 1 Total Cost $1.546 Response Time (avg) 7.47s

Filter models