Did not follow instructions Failure Ranking

AI BENCHY Failures

See which AI models run into Did not follow instructions most often, so you can spot reliability risks before choosing one. Sort by: Score ↑.

Models Shown

Total Failures

215

Most Affected Model

Categories

In category Puzzle Solving78 In category General Intelligence74 In category Anti-AI Tricks30 In category Coding14 In category Instructions following11 In category Tool Calling6 In category Combined1 In category Domain specific1

Rank	Model	Company	Did not follow instructions Count	Score	Tests Correct	Response Time (avg)
#92	Laguna M.1 medium	Poolside	1	6.4	9/19	14.7s
#90	Gemini 3.1 Flash Lite none	Google	1	6.4	9/21	1.06s
#88	Qwen3.7 Plus none	Qwen	1	6.4	10/21	2.85s
#87	Gemini 3.1 Flash Lite minimal	Google	3	6.4	10/21	1.33s
#86	Grok 4.1 Fast medium	X AI	4	6.5	9/19	23.8s
#85	Gemma 4 31B none	Google	1	6.5	10/21	4.05s
#84	Grok 4.20 Multi Agent Beta medium	X AI	2	6.6	8/18	9.69s
#83	Step 3.5 Flash none	Stepfun	1	6.6	6/12	39.0s
#81	Mercury 2 medium	Inception	3	6.6	10/21	2.24s
#80	Mimo V2 Omni medium	Xiaomi	2	6.7	10/21	41.2s
#79	Hunter Alpha medium	OpenRouter	2	6.7	8/18	10.3s
#78	Qwen3.6 27B medium	Qwen	1	6.8	10/21	59.7s
#77	Claude Sonnet 4.6 none	Anthropic	1	6.8	11/21	5.04s
#76	Kimi K2.5 medium	Moonshot AI	2	6.8	10/21	98.4s
#75	Ring-2.6-1T medium	Inclusionai	2	6.9	11/21	61.3s

←

1 4 5 6 8

→

Did not follow instructions Failures