Coding Model Ranking

AI BENCHY Category

See which AI models perform best on Coding, which ones stay reliable, and where the biggest gaps appear.

Models Shown

Average Coding Score

6.1

Best Model

Failure Reasons

With failure reason Wrong answer120 With failure reason No answer18 With failure reason Did not follow instructions16 With failure reason Timed out12 With failure reason Extra formatting7 With failure reason API error6

Rank	Model	Company	Coding Score	Score	Tests Correct	Response Time (avg)
#69	Claude Sonnet 4.6 none	Anthropic	6.8	7.0	1/2	6.73s
#76	Gemma 4 31B none	Google	6.8	6.7	1/2	14.8s
#78	Gemini 3.1 Flash Lite minimal	Google	6.8	6.7	1/2	951ms
#85	Gemini 3.1 Flash Lite none	Google	6.8	6.6	1/2	1.13s
#86	GPT-5.5 none	OpenAI	6.8	6.5	1/2	1.52s
#92	Gemini 2.5 Flash none	Google	6.8	6.2	1/2	810ms
#97	Qwen3.5-Flash none	Qwen	6.8	5.9	1/2	993ms
#98	GLM 5V Turbo none	Z.ai	6.8	5.9	1/2	3.77s
#102	Qwen3.5-35B-A3B none	Qwen	6.8	5.8	1/2	1.72s
#104	Qwen3.6 27B none	Qwen	6.8	5.8	1/2	5.75s
#107	MiMo-V2-Pro none	Xiaomi	6.8	5.7	1/2	2.65s
#112	GPT-5.4 none	OpenAI	6.8	5.6	1/2	1.99s
#126	Kimi K2.5 none	Moonshot AI	6.8	5.3	1/2	36.0s
#136	GPT-5.4 Mini none	OpenAI	6.8	4.9	1/2	1.01s
#137	Qwen3.6 35B A3B none	Qwen	6.8	4.9	1/2	12.3s

Coding Ranking