Instructions following x Extra formatting Ranking

See which AI models are most likely to hit Extra formatting on Instructions following, so you can spot weak points faster. Sort by: Failure Count ↑.

Models Shown

Total Failures

Most Affected Model

Failure Reasons

Wrong answer82 Did not follow instructions48 Extra formatting7 Timed out5 API error2 No answer2

Categories

Anti-AI Tricks29 Coding26 Domain specific23 Puzzle Solving14 Data parsing and extraction8 Instructions following7 Combined1

7/7

Rank	Model	Company	Extra formatting Count	Category Score	Total Cost	Tests Correct	Response Time (avg)
#114	DeepSeek V4 Flash 0731 medium	DeepSeek	1	8.5	$0.066	1/2	2.42s
Total Tests 2 Wrong Tests 1 Total Cost $0.066 Response Time (avg) 2.42s
#136	Claude Fable 5.1 low	Anthropic	1	6.5	$2.565	1/2	5.24s
Total Tests 2 Wrong Tests 1 Total Cost $2.565 Response Time (avg) 5.24s
#181	Claude Sonnet 5 none	Anthropic	1	6.4	$0.548	1/2	2.58s
Total Tests 2 Wrong Tests 1 Total Cost $0.548 Response Time (avg) 2.58s
#233	DeepSeek V4 Flash 0423 none	DeepSeek	1	6.5	$0.037	1/2	17.5s
Total Tests 2 Wrong Tests 1 Total Cost $0.037 Response Time (avg) 17.5s
#243	Seed-2.0-Code none	Bytedance Seed	1	6.2	$0.151	1/2	1.80s
Total Tests 2 Wrong Tests 1 Total Cost $0.151 Response Time (avg) 1.80s
#291	Granite 4.2 8B none	IBM Granite	1	3.0	$0.026	0/2	79.5s
Total Tests 2 Wrong Tests 2 Total Cost $0.026 Response Time (avg) 79.5s
#295	Hy3 preview none	Tencent	1	6.3	$0.007	1/2	13.0s
Total Tests 2 Wrong Tests 1 Total Cost $0.007 Response Time (avg) 13.0s

Filter models