Skip to main content
benchlm.ai

Artificial Analysis IFBench (AA-IFBench)

A display-only Artificial Analysis IFBench score.

Benchmark score on AA-IFBench — May 21, 2026

BenchLM mirrors the published score view for AA-IFBench. Grok 4.3 leads the public snapshot at 81.3% , followed by Qwen3.7 Max (80.5%) and MiMo-V2.5-Pro (79.9%). BenchLM does not use these results to rank models overall.

1Closed

Grok 4.3

xAI

81.3%
Overall —Context 1M
2Closed

Qwen3.7 Max

Alibaba

80.5%
Overall 92Context 1M
3Closed

MiMo-V2.5-Pro

Xiaomi

79.9%
Overall —Context 1M
119 modelsInstruction FollowingCurrentDisplay onlyUpdated May 21, 2026

The published AA-IFBench snapshot is tightly clustered at the top: Grok 4.3 sits at 81.3%, while the third row is only 1.4 points behind. The broader top-10 spread is 4.8 points, so many of the published scores sit in a relatively narrow band.

119 models have been evaluated on AA-IFBench. The benchmark falls in the Instruction Following category. This category carries a 5% weight in BenchLM.ai's overall scoring system. AA-IFBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About AA-IFBench

Year

2026

Tasks

Verifiable instruction constraints

Format

Constraint satisfaction accuracy

Difficulty

Instruction precision

BenchLM stores the Artificial Analysis IFBench result separately from the weighted IFBench lane so AA refreshes remain display-only.

Artificial Analysis IFBench Benchmark Leaderboard

BenchLM freshness & provenance

Version

AA-IFBench 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Benchmark score table (119 models)

1
Grok 4.3
81.3%
2
Qwen3.7 Max
80.5%
3
MiMo-V2.5-Pro
79.9%
4
DeepSeek V4 Flash (Max)
79.2%
5
Qwen3.5 397B (Reasoning)
78.8%
6
GPT-5.2-Codex
77.6%
7
Gemini 3.1 Flash-Lite
77.2%
8
Gemini 3.1 Pro
77.1%
9
Qwen 3.6 Max (preview)
76.6%
10
DeepSeek V4 Pro (Max)
76.5%
11
Gemini 3.5 Flash
76.3%
12
GLM-5.1
76.3%
13
Kimi K2.6
76.0%
14
GPT-5.5
75.9%
15
Muse Spark
75.9%
16
GPT-5.4 nano
75.9%
17
Qwen3.5-122B-A10B
75.7%
18
MiniMax M2.7
75.7%
19
Qwen3.5-27B
75.6%
20
Gemma 4 31B
75.6%
21
GPT-5.3 Codex
75.4%
22
GPT-5.2
75.4%
23
Qwen3.6 Plus
75.2%
24
GPT-5.4
73.9%
25
Command A+
73.9%
26
DeepSeek V4 Flash (High)
73.5%
27
GPT-5.4 mini
73.3%
28
GLM-5-Turbo
73.2%
29
GPT-5 (high)
73.1%
30
GPT-5.1
72.9%
31
Qwen3.5-35B-A3B
72.5%
32
Gemma 4 26B A4B
72.4%
33
GLM-5 (Reasoning)
72.3%
34
GLM-5
72.3%
35
o3
71.4%
36
DeepSeek V4 Pro (High)
71.3%
37
GPT-5 (medium)
70.6%
38
Gemini 3 Pro
70.4%
39
o1
70.3%
40
Kimi K2.5 (Reasoning)
70.2%
41
Kimi K2.5
70.2%
42
GPT-5.1-Codex-Max
70.0%
43
GPT-5.1-Codex
70.0%
44
GPT-OSS 120B
69.0%
45
MiMo-V2-Pro
68.8%
46
Mistral Medium 3.5 128B
68.8%
47
GLM-4.7
67.9%
48
Qwen3.6-27B
67.6%
49
GPT-OSS 20B
65.1%
50
K-Exaone
64.7%
51
Qwen3.6-35B-A3B
64.4%
52
Nemotron 3 Nano Omni 30B A3B
63.2%
53
Hy3 Preview
63.1%
54
GLM-5V-Turbo
61.1%
55
Claude Opus 4.7 (Adaptive)
58.6%
56
Claude Opus 4.5 Thinking
58.0%
57
Ling 2.6 Flash
57.4%
58
Trinity-Large-Preview
56.3%
59
Trinity-Large-Thinking
56.3%
60
Claude 4.1 Opus Thinking
55.4%
61
Gemini 3 Flash
55.1%
62
Grok 4
53.7%
63
MiMo-V2-Omni
53.5%
64
Claude Opus 4.6 (Adaptive)
53.1%
65
Grok 4.1 Fast (Reasoning)
52.7%
66
Qwen3.5 397B
51.6%
67
Grok 4 Fast (Reasoning)
50.5%
68
Grok 4.20
49.3%
69
DeepSeek V3.2
49.0%
70
Gemini 2.5 Pro
48.7%
71
Mistral Small 4 (Reasoning)
48.2%
72
Mistral Small 4
48.2%
73
Grok 3 Mini
45.9%
74
Claude 4 Sonnet
45.4%
75
Claude Opus 4.6
44.6%
76
Gemma 4 E4B
44.2%
77
Qwen3 Max
44.1%
78
Claude Opus 4.7
43.6%
79
Claude Opus 4.5
43.0%
80
GPT-4.1
43.0%
81
Llama 4 Maverick
43.0%
82
Kimi K2
41.5%
83
DeepSeek V3.1 (Reasoning)
41.5%
84
Grok Code Fast 1
41.4%
85
Claude Sonnet 4.6
41.2%
86
MiMo-V2-Flash
39.9%
87
DeepSeek-R1
39.6%
88
Llama 4 Scout
39.5%
89
Mistral Medium 3
39.3%
90
Llama 3.1 405B
39.0%
91
Gemini 2.5 Flash
39.0%
92
GPT-4.1 mini
38.3%
93
Nemotron Ultra 253B
38.2%
94
Nova Pro
38.1%
95
Gemma 4 E2B
38.0%
96
DeepSeek V3.1
37.8%
97
GLM-4.5-Air
37.6%
98
Nemotron 3 Nano 30B
37.5%
99
GLM-4.6
36.7%
100
Grok 4.1 Fast
36.5%
101
Mistral Large 3
36.2%
102
Claude 3 Haiku
36.1%
103
DeepSeek V3
34.8%
104
Sarvam 105B
34.4%
105
GPT-4o
34.3%
106
Solar Pro 2
33.7%
107
Exaone 4.0 32B
33.5%
108
GPT-4.1 nano
32.0%
109
Gemma 3 27B
31.8%
110
Mistral Large 2
31.2%
111
GPT-4o mini
31.0%
112
Sarvam 30B
26.5%
113
Granite-4.0-H-1B
26.2%
114
Exaone 4.0 1.2B
25.3%
115
Phi-4
23.5%
116
DeepSeek R1 Distill Qwen 32B
22.9%
117
Granite-4.0-1B
20.5%
118
Granite-4.0-H-350M
17.6%
119
Granite-4.0-350M
15.9%

FAQ

What does AA-IFBench measure?

A display-only Artificial Analysis IFBench score.

Which model scores highest on AA-IFBench?

Grok 4.3 by xAI currently leads with a score of 81.3% on AA-IFBench.

How many models are evaluated on AA-IFBench?

119 AI models have been evaluated on AA-IFBench on BenchLM.

Compare Top Models on AA-IFBench

Grok 4.3 vs Qwen3.7 MaxQwen3.7 Max vs MiMo-V2.5-ProMiMo-V2.5-Pro vs DeepSeek V4 Flash (Max)DeepSeek V4 Flash (Max) vs Qwen3.5 397B (Reasoning)
Last updated: May 21, 2026 · BenchLM version AA-IFBench 2026

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.