# Artificial Analysis IFBench (AA-IFBench)

> A display-only Artificial Analysis IFBench score.

Canonical page: https://benchlm.ai/benchmarks/aaifbench

- Category: [Instruction Following](/instruction-following)
- Last updated: September 10, 2026

## About AA-IFBench

- Year: 2026
- Tasks: Verifiable instruction constraints
- Format: Constraint satisfaction accuracy
- Difficulty: Instruction precision
- Paper: [Artificial Analysis IFBench Benchmark Leaderboard](https://artificialanalysis.ai/evaluations/ifbench)

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

AA-IFBench is currently weighted in BenchLM's scoring formula. The Instruction Following category carries 5% of the overall score, and AA-IFBench contributes 30% of that category score.

## Leaderboard (130 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [MiniMax M3](/models/minimax-m3) | MiniMax | 82.9% |
| 2 | [Nemotron 3 Ultra](/models/nemotron-3-ultra) | NVIDIA | 81.4% |
| 3 | [Grok 4.3](/models/grok-4-3) | xAI | 81.3% |
| 4 | [Qwen3.7 Max](/models/qwen3-7-max) | Alibaba | 80.5% |
| 5 | [MiMo-V2.5-Pro](/models/mimo-v2-5-pro) | Xiaomi | 79.9% |
| 6 | [Qwen3.7 Plus](/models/qwen3-7-plus) | Alibaba | 78.0% |
| 7 | [GPT-5.2-Codex](/models/gpt-5-2-codex) | OpenAI | 77.6% |
| 8 | [Gemini 3.1 Pro](/models/gemini-3-1-pro) | Google | 77.1% |
| 9 | [Qwen 3.6 Max (preview)](/models/qwen3-6-max-preview) | Alibaba | 76.6% |
| 10 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | DeepSeek | 76.5% |
| 11 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | Google | 76.3% |
| 12 | [GLM-5.1](/models/glm-5-1) | Z.AI | 76.3% |
| 13 | [Kimi K2.6](/models/kimi-2-6) | Moonshot AI | 76.0% |
| 14 | [GPT-5.5](/models/gpt-5-5) | OpenAI | 75.9% |
| 15 | [Muse Spark](/models/muse-spark) | Meta | 75.9% |
| 16 | [GPT-5.4 nano](/models/gpt-5-4-nano) | OpenAI | 75.9% |
| 17 | [Qwen3.5-122B-A10B](/models/qwen3-5-122b-a10b) | Alibaba | 75.7% |
| 18 | [MiniMax M2.7](/models/minimax-m2-7) | MiniMax | 75.7% |
| 19 | [Qwen3.5-27B](/models/qwen3-5-27b) | Alibaba | 75.6% |
| 20 | [Gemma 4 31B](/models/gemma-4-31b) | Google | 75.6% |
| 21 | [GPT-5.2](/models/gpt-5-2) | OpenAI | 75.4% |
| 22 | [GPT-5.3 Codex](/models/gpt-5-3-codex) | OpenAI | 75.4% |
| 23 | [Qwen3.6 Plus](/models/qwen3-6-plus) | Alibaba | 75.2% |
| 24 | [GPT-5.4](/models/gpt-5-4) | OpenAI | 73.9% |
| 25 | [Command A+](/models/command-a-plus) | Cohere | 73.9% |
| 26 | [Gemma 4 12B](/models/gemma-4-12b) | Google | 73.5% |
| 27 | [GLM-5.2](/models/glm-5-2) | Z.AI | 73.3% |
| 28 | [GPT-5.4 mini](/models/gpt-5-4-mini) | OpenAI | 73.3% |
| 29 | [GLM-5-Turbo](/models/glm-5-turbo) | Z.AI | 73.2% |
| 30 | [GPT-5 (high)](/models/gpt-5-high) | OpenAI | 73.1% |
| 31 | [GPT-5.1](/models/gpt-5-1) | OpenAI | 72.9% |
| 32 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | OpenAI | 72.7% |
| 33 | [Qwen3.5-35B-A3B](/models/qwen3-5-35b-a3b) | Alibaba | 72.5% |
| 34 | [Gemma 4 26B A4B](/models/gemma-4-26b-a4b) | Google | 72.4% |
| 35 | [GLM-5](/models/glm-5) | Z.AI | 72.3% |
| 36 | [Nemotron 3 Super 100B](/models/nemotron-3-super-100b) | NVIDIA | 71.5% |
| 37 | [o3](/models/o3) | OpenAI | 71.4% |
| 38 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | OpenAI | 71.2% |
| 39 | [Solar Pro 3](/models/solar-pro-3) | Upstage | 71.2% |
| 40 | [Nemotron 3 Nano 30B](/models/nemotron-3-nano-30b) | NVIDIA | 71.1% |
| 41 | [GPT-5 (medium)](/models/gpt-5-medium) | OpenAI | 70.6% |
| 42 | [Gemini 3 Pro](/models/gemini-3-pro) | Google | 70.4% |
| 43 | [o1](/models/o1) | OpenAI | 70.3% |
| 44 | [Kimi K2.5 (Reasoning)](/models/kimi-k2-5-reasoning) | Moonshot AI | 70.2% |
| 45 | [Kimi K2.5](/models/kimi-k2-5) | Moonshot AI | 70.2% |
| 46 | [GPT-5.1-Codex-Max](/models/gpt-5-1-codex-max) | OpenAI | 70.0% |
| 47 | [GPT-5.1-Codex](/models/gpt-5-1-codex) | OpenAI | 70.0% |
| 48 | [GPT-OSS 120B](/models/gpt-oss-120b) | OpenAI | 69.0% |
| 49 | [MiMo-V2-Pro](/models/mimo-v2-pro) | Xiaomi | 68.8% |
| 50 | [Mistral Medium 3.5 128B](/models/mistral-medium-3-5-128b) | Mistral | 68.8% |
| 51 | [GLM-4.7](/models/glm-4-7) | Z.AI | 67.9% |
| 52 | [Qwen3.6-27B](/models/qwen3-6-27b) | Alibaba | 67.6% |
| 53 | [Step 3.7 Flash](/models/step-3-7-flash) | StepFun | 67.3% |
| 54 | [GPT-OSS 20B](/models/gpt-oss-20b) | OpenAI | 65.1% |
| 55 | [K-Exaone](/models/k-exaone) | LG AI Research | 64.7% |
| 56 | [Qwen3.6-35B-A3B](/models/qwen3-6-35b-a3b) | Alibaba | 64.4% |
| 57 | [Claude Fable 5](/models/claude-fable) | Anthropic | 63.5% |
| 58 | [Nemotron 3 Nano Omni 30B A3B](/models/nemotron-3-nano-omni-30b-a3b) | NVIDIA | 63.2% |
| 59 | [Kimi K2.7 Code](/models/kimi-k2-7-code) | Moonshot AI | 63.1% |
| 60 | [Claude Opus 4.8](/models/claude-opus-4-8) | Anthropic | 62.2% |
| 61 | [GLM-5V-Turbo](/models/glm-5v-turbo) | Z.AI | 61.1% |
| 62 | [Claude Opus 4.7 (Adaptive)](/models/claude-opus-4-7-adaptive) | Anthropic | 58.6% |
| 63 | [Claude Opus 4.5 Thinking](/models/claude-opus-4-5-thinking) | Anthropic | 58.0% |
| 64 | [North Mini Code](/models/north-mini-code-1-0) | Cohere | 57.6% |
| 65 | [Ling 2.6 Flash](/models/ling-2-6-flash) | InclusionAI | 57.4% |
| 66 | [Trinity-Large-Preview](/models/trinity-large-preview) | Arcee AI | 56.3% |
| 67 | [Trinity-Large-Thinking](/models/trinity-large-thinking) | Arcee AI | 56.3% |
| 68 | [LFM2.5-8B-A1B](/models/lfm2-5-8b-a1b) | LiquidAI | 55.6% |
| 69 | [Claude 4.1 Opus Thinking](/models/claude-4-1-opus-thinking) | Anthropic | 55.4% |
| 70 | [Gemini 3 Flash](/models/gemini-3-flash) | Google | 55.1% |
| 71 | [Grok 4](/models/grok-4) | xAI | 53.7% |
| 72 | [MiMo-V2-Omni](/models/mimo-v2-omni) | Xiaomi | 53.5% |
| 73 | [Claude Opus 4.6 (Adaptive)](/models/claude-opus-4-6-thinking) | Anthropic | 53.1% |
| 74 | [Grok 4.1 Fast (Reasoning)](/models/grok-4-1-fast-reasoning) | xAI | 52.7% |
| 75 | [Qwen3.5 397B (Reasoning)](/models/qwen3-5-397b-reasoning) | Alibaba | 51.6% |
| 76 | [Qwen3.5 397B](/models/qwen3-5-397b) | Alibaba | 51.6% |
| 77 | [Grok 4 Fast (Reasoning)](/models/grok-4-fast-reasoning) | xAI | 50.5% |
| 78 | [DeepSeek V3.2](/models/deepseek-v3-2) | DeepSeek | 49.0% |
| 79 | [Gemini 2.5 Pro](/models/gemini-2-5-pro) | Google | 48.7% |
| 80 | [Mistral Small 4](/models/mistral-small-4) | Mistral | 48.2% |
| 81 | [Mistral Small 4 (Reasoning)](/models/mistral-small-4-reasoning) | Mistral | 48.2% |
| 82 | [Ultravox v0.6 Llama 3.3 70B](/models/ultravox-v0-6-llama-3-3-70b) | Fixie AI | 47.1% |
| 83 | [Claude 4 Sonnet](/models/claude-4-sonnet) | Anthropic | 45.4% |
| 84 | [Claude Opus 4.6](/models/claude-opus-4-6) | Anthropic | 44.6% |
| 85 | [Gemma 4 E4B](/models/gemma-4-e4b) | Google | 44.2% |
| 86 | [Qwen3 Max](/models/qwen3-max) | Alibaba | 44.1% |
| 87 | [Claude Opus 4.7](/models/claude-opus-4-7) | Anthropic | 43.6% |
| 88 | [Qwen3-Omni-30B-A3B-Thinking](/models/qwen3-omni-30b-a3b-thinking) | Alibaba | 43.4% |
| 89 | [Claude Opus 4.5](/models/claude-opus-4-5) | Anthropic | 43.0% |
| 90 | [GPT-4.1](/models/gpt-4-1) | OpenAI | 43.0% |
| 91 | [Llama 4 Maverick](/models/llama-4-maverick) | Meta | 43.0% |
| 92 | [DeepSeek V3.1 (Reasoning)](/models/deepseek-v3-1-reasoning) | DeepSeek | 41.5% |
| 93 | [Kimi K2](/models/kimi-k2) | Moonshot AI | 41.5% |
| 94 | [Grok Code Fast 1](/models/grok-code-fast-1) | xAI | 41.4% |
| 95 | [Claude Sonnet 4.6](/models/claude-sonnet-4-6) | Anthropic | 41.2% |
| 96 | [MiMo-V2-Flash](/models/mimo-v2-flash) | Xiaomi | 39.9% |
| 97 | [DeepSeek-R1](/models/deepseek-r1) | DeepSeek | 39.6% |
| 98 | [Llama 4 Scout](/models/llama-4-scout) | Meta | 39.5% |
| 99 | [Mistral Medium 3](/models/mistral-medium-3) | Mistral | 39.3% |
| 100 | [Gemini 2.5 Flash](/models/gemini-2-5-flash) | Google | 39.0% |
| 101 | [Llama 3.1 405B](/models/llama-3-1-405b) | Meta | 39.0% |
| 102 | [GPT-4.1 mini](/models/gpt-4-1-mini) | OpenAI | 38.3% |
| 103 | [Nemotron Ultra 253B](/models/nemotron-ultra-253b) | NVIDIA | 38.2% |
| 104 | [Nova Pro](/models/nova-pro) | Amazon | 38.1% |
| 105 | [Gemma 4 E2B](/models/gemma-4-e2b) | Google | 38.0% |
| 106 | [DeepSeek V3.1](/models/deepseek-v3-1) | DeepSeek | 37.8% |
| 107 | [GLM-4.5-Air](/models/glm-4-5-air) | Z.AI | 37.6% |
| 108 | [GLM-4.6](/models/glm-4-6) | Z.AI | 36.7% |
| 109 | [Grok 4.1 Fast](/models/grok-4-1-fast) | xAI | 36.5% |
| 110 | [Mistral Large 3](/models/mistral-large-3) | Mistral | 36.2% |
| 111 | [Claude 3 Haiku](/models/claude-3-haiku) | Anthropic | 36.1% |
| 112 | [DeepSeek V3](/models/deepseek-v3) | DeepSeek | 34.8% |
| 113 | [Sarvam 105B](/models/sarvam-105b) | Sarvam | 34.4% |
| 114 | [GPT-4o](/models/gpt-4o) | OpenAI | 34.3% |
| 115 | [Solar Pro 2](/models/solar-pro-2) | Upstage | 33.7% |
| 116 | [Exaone 4.0 32B](/models/exaone-4-0-32b) | LG AI Research | 33.5% |
| 117 | [LFM2.5-VL-1.6B-Extract](/models/lfm2-5-vl-1-6b-extract) | LiquidAI | 33.1% |
| 118 | [GPT-4.1 nano](/models/gpt-4-1-nano) | OpenAI | 32.0% |
| 119 | [Gemma 3 27B](/models/gemma-3-27b) | Google | 31.8% |
| 120 | [Qwen3-Omni-30B-A3B-Instruct](/models/qwen3-omni-30b-a3b-instruct) | Alibaba | 31.2% |
| 121 | [Mistral Large 2](/models/mistral-large-2) | Mistral | 31.2% |
| 122 | [GPT-4o mini](/models/gpt-4o-mini) | OpenAI | 31.0% |
| 123 | [Sarvam 30B](/models/sarvam-30b) | Sarvam | 26.5% |
| 124 | [Granite-4.0-H-1B](/models/granite-4-0-h-1b) | IBM | 26.2% |
| 125 | [Exaone 4.0 1.2B](/models/exaone-4-0-1-2b) | LG AI Research | 25.3% |
| 126 | [Phi-4](/models/phi-4) | Microsoft | 23.5% |
| 127 | [DeepSeek R1 Distill Qwen 32B](/models/deepseek-r1-distill-qwen-32b) | DeepSeek | 22.9% |
| 128 | [Granite-4.0-1B](/models/granite-4-0-1b) | IBM | 20.5% |
| 129 | [Granite-4.0-H-350M](/models/granite-4-0-h-350m) | IBM | 17.6% |
| 130 | [Granite-4.0-350M](/models/granite-4-0-350m) | IBM | 15.9% |

## FAQ

### What does AA-IFBench measure?

A display-only Artificial Analysis IFBench score.

### Which model scores highest on AA-IFBench?

MiniMax M3 by MiniMax currently leads with a score of 82.9% on AA-IFBench.

### How many models are evaluated on AA-IFBench?

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

### Does AA-IFBench affect BenchLM's overall score?

Yes. AA-IFBench is a weighted benchmark inside the Instruction Following category, which carries 5% of BenchLM's overall score. AA-IFBench itself contributes 30% of that category score.

## Compare Top Models on AA-IFBench

- [MiniMax M3 vs Nemotron 3 Ultra](/compare/minimax-m3-vs-nemotron-3-ultra)
- [Nemotron 3 Ultra vs Grok 4.3](/compare/grok-4-3-vs-nemotron-3-ultra)
- [Grok 4.3 vs Qwen3.7 Max](/compare/grok-4-3-vs-qwen3-7-max)
- [Qwen3.7 Max vs MiMo-V2.5-Pro](/compare/mimo-v2-5-pro-vs-qwen3-7-max)