aa If Bench

116 models evaluated

#	Model	Provider	Type	Score
1	Grok 4.3	xAI	Closed	81.3%
2	Qwen3.7 Max	Alibaba	Closed	80.5%
3	MiMo-V2.5-Pro	Xiaomi	Closed	79.9%
4	DeepSeek V4 Flash (Max)	DeepSeek	Open	79.2%
5	Qwen3.5 397B (Reasoning)	Alibaba	Open	78.8%
6	GPT-5.2-Codex	OpenAI	Closed	77.6%
7	Gemini 3.1 Flash-Lite	Google	Closed	77.2%
8	Gemini 3.1 Pro	Google	Closed	77.1%
9	Qwen 3.6 Max (preview)	Alibaba	Closed	76.6%
10	DeepSeek V4 Pro (Max)	DeepSeek	Open	76.5%
11	Gemini 3.5 Flash	Google	Closed	76.3%
12	GLM-5.1	Z.AI	Open	76.3%
13	Kimi K2.6	Moonshot AI	Open	76.0%
14	GPT-5.5	OpenAI	Closed	75.9%
15	Muse Spark	Meta	Closed	75.9%
16	GPT-5.4 nano	OpenAI	Closed	75.9%
17	Qwen3.5-122B-A10B	Alibaba	Open	75.7%
18	MiniMax M2.7	MiniMax	Open	75.7%
19	Qwen3.5-27B	Alibaba	Open	75.6%
20	Gemma 4 31B	Google	Open	75.6%
21	GPT-5.3 Codex	OpenAI	Closed	75.4%
22	GPT-5.2	OpenAI	Closed	75.4%
23	Qwen3.6 Plus	Alibaba	Closed	75.2%
24	GPT-5.4	OpenAI	Closed	73.9%
25	Command A+	Cohere	Open	73.9%
26	DeepSeek V4 Flash (High)	DeepSeek	Open	73.5%
27	GPT-5.4 mini	OpenAI	Closed	73.3%
28	GLM-5-Turbo	Z.AI	Closed	73.2%
29	GPT-5 (high)	OpenAI	Closed	73.1%
30	GPT-5.1	OpenAI	Closed	72.9%
31	Qwen3.5-35B-A3B	Alibaba	Open	72.5%
32	Gemma 4 26B A4B	Google	Open	72.4%
33	GLM-5	Z.AI	Open	72.3%
34	o3	OpenAI	Closed	71.4%
35	DeepSeek V4 Pro (High)	DeepSeek	Open	71.3%
36	GPT-5 (medium)	OpenAI	Closed	70.6%
37	Gemini 3 Pro	Google	Closed	70.4%
38	o1	OpenAI	Closed	70.3%
39	Kimi K2.5 (Reasoning)	Moonshot AI	Closed	70.2%
40	Kimi K2.5	Moonshot AI	Open	70.2%
41	GPT-5.1-Codex-Max	OpenAI	Closed	70.0%
42	GPT-5.1-Codex	OpenAI	Closed	70.0%
43	GPT-OSS 120B	OpenAI	Open	69.0%
44	MiMo-V2-Pro	Xiaomi	Closed	68.8%
45	Mistral Medium 3.5 128B	Mistral	Open	68.8%
46	GLM-4.7	Z.AI	Open	67.9%
47	Qwen3.6-27B	Alibaba	Open	67.6%
48	GPT-OSS 20B	OpenAI	Open	65.1%
49	K-Exaone	LG AI Research	Closed	64.7%
50	Qwen3.6-35B-A3B	Alibaba	Open	64.4%
51	Nemotron 3 Nano Omni 30B A3B	NVIDIA	Open	63.2%
52	Hy3 Preview	Tencent	Open	63.1%
53	GLM-5V-Turbo	Z.AI	Closed	61.1%
54	Claude Opus 4.7 (Adaptive)	Anthropic	Closed	58.6%
55	Claude Opus 4.5 Thinking	Anthropic	Closed	58.0%
56	Ling 2.6 Flash	InclusionAI	Open	57.4%
57	Trinity-Large-Preview	Arcee AI	Open	56.3%
58	Trinity-Large-Thinking	Arcee AI	Open	56.3%
59	Claude 4.1 Opus Thinking	Anthropic	Closed	55.4%
60	Gemini 3 Flash	Google	Closed	55.1%
61	Grok 4	xAI	Closed	53.7%
62	MiMo-V2-Omni	Xiaomi	Closed	53.5%
63	Claude Opus 4.6 (Adaptive)	Anthropic	Closed	53.1%
64	Grok 4.1 Fast (Reasoning)	xAI	Closed	52.7%
65	Qwen3.5 397B	Alibaba	Open	51.6%
66	Grok 4 Fast (Reasoning)	xAI	Closed	50.5%
67	DeepSeek V3.2	DeepSeek	Open	49.0%
68	Gemini 2.5 Pro	Google	Closed	48.7%
69	Mistral Small 4 (Reasoning)	Mistral	Open	48.2%
70	Mistral Small 4	Mistral	Open	48.2%
71	Claude 4 Sonnet	Anthropic	Closed	45.4%
72	Claude Opus 4.6	Anthropic	Closed	44.6%
73	Gemma 4 E4B	Google	Open	44.2%
74	Qwen3 Max	Alibaba	Closed	44.1%
75	Claude Opus 4.7	Anthropic	Closed	43.6%
76	Claude Opus 4.5	Anthropic	Closed	43.0%
77	GPT-4.1	OpenAI	Closed	43.0%
78	Llama 4 Maverick	Meta	Open	43.0%
79	Kimi K2	Moonshot AI	Closed	41.5%
80	DeepSeek V3.1 (Reasoning)	DeepSeek	Open	41.5%
81	Grok Code Fast 1	xAI	Closed	41.4%
82	Claude Sonnet 4.6	Anthropic	Closed	41.2%
83	MiMo-V2-Flash	Xiaomi	Open	39.9%
84	DeepSeek-R1	DeepSeek	Open	39.6%
85	Llama 4 Scout	Meta	Open	39.5%
86	Mistral Medium 3	Mistral	Closed	39.3%
87	Llama 3.1 405B	Meta	Open	39.0%
88	Gemini 2.5 Flash	Google	Closed	39.0%
89	GPT-4.1 mini	OpenAI	Closed	38.3%
90	Nemotron Ultra 253B	NVIDIA	Open	38.2%
91	Nova Pro	Amazon	Closed	38.1%
92	Gemma 4 E2B	Google	Open	38.0%
93	DeepSeek V3.1	DeepSeek	Open	37.8%
94	GLM-4.5-Air	Z.AI	Closed	37.6%
95	Nemotron 3 Nano 30B	NVIDIA	Open	37.5%
96	GLM-4.6	Z.AI	Open	36.7%
97	Grok 4.1 Fast	xAI	Closed	36.5%
98	Mistral Large 3	Mistral	Closed	36.2%
99	Claude 3 Haiku	Anthropic	Closed	36.1%
100	DeepSeek V3	DeepSeek	Open	34.8%
101	Sarvam 105B	Sarvam	Open	34.4%
102	GPT-4o	OpenAI	Closed	34.3%
103	Solar Pro 2	Upstage	Closed	33.7%
104	Exaone 4.0 32B	LG AI Research	Open	33.5%
105	GPT-4.1 nano	OpenAI	Closed	32.0%
106	Gemma 3 27B	Google	Open	31.8%
107	Mistral Large 2	Mistral	Closed	31.2%
108	GPT-4o mini	OpenAI	Closed	31.0%
109	Sarvam 30B	Sarvam	Open	26.5%
110	Granite-4.0-H-1B	IBM	Open	26.2%
111	Exaone 4.0 1.2B	LG AI Research	Open	25.3%
112	Phi-4	Microsoft	Open	23.5%
113	DeepSeek R1 Distill Qwen 32B	DeepSeek	Open	22.9%
114	Granite-4.0-1B	IBM	Open	20.5%
115	Granite-4.0-H-350M	IBM	Open	17.6%
116	Granite-4.0-350M	IBM	Open	15.9%