Instructions following Model Ranking

See which AI models perform best on Instructions following, which ones stay reliable, and where the biggest gaps appear. Sort by: Response Time (avg) ↑.

Models Shown

Average Instructions following Score

8.6

Best Model

Granite 4.1 8B 3.6

Failure Reasons

With failure reason Wrong answer61 With failure reason Did not follow instructions19 With failure reason Extra formatting3 With failure reason No answer2 With failure reason API error1 With failure reason Timed out1

216/216

Rank	Model	Company	Instructions following Score	Score	Total Cost	Tests Correct	Response Time (avg)
#141	Hy3 preview high	Tencent	10.0	5.9	$0.048	2/2	34.4s
Total Tests 2 Wrong Tests 0 Total Cost $0.048 Response Time (avg) 34.4s
#80	DeepSeek V3.2 medium	DeepSeek	10.0	7.0	$0.078	2/2	35.8s
Total Tests 2 Wrong Tests 0 Total Cost $0.078 Response Time (avg) 35.8s
#105	Qwen3.6 27B medium	Qwen	10.0	6.5	$0.779	2/2	38.0s
Total Tests 2 Wrong Tests 0 Total Cost $0.779 Response Time (avg) 38.0s
#120	Qwen3.5-Flash medium	Qwen	10.0	6.2	$0.139	2/2	63.5s
Total Tests 2 Wrong Tests 0 Total Cost $0.139 Response Time (avg) 63.5s
#169	Gemini 3.1 Flash Lite Preview high	Google	9.8	5.3	$2.310	2/2	64.0s
Total Tests 2 Wrong Tests 0 Total Cost $2.310 Response Time (avg) 64.0s
#81	Kimi K2.5 medium	Moonshot AI	10.0	7.0	$0.600	2/2	92.5s
Total Tests 2 Wrong Tests 0 Total Cost $0.600 Response Time (avg) 92.5s

Instructions following Ranking

Filter models

Top Models by Instructions following Score

Instructions following Score vs Total Cost

Top Models by Response Time (avg)