IFBench (instruction-following benchmark, harder successor to IFEval) — Real published benchmark; aa_ifbench raw confirms AA measures this same public benchmark.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Inkling | Thinking Machines | 83.4 | 2 | 2026-07-16 |
| 2 | Grok 4.3 | SpaceXAI | 83.3 | 5 | 2026-05-23 |
| 3 | Grok 4.20 | SpaceXAI | 82.9 | 1 | 2026-05-02 |
| 4 | MiniMax M3 | MiniMax | 82.9 | 2 | 2026-07-07 |
| 5 | Qwen3.8 Max Preview | Alibaba | 82.8 | 2 | 2026-08-23 |
| 6 | Inkling-Small | Thinking Machines | 82.2 | 3 | 2026-08-24 |
| 7 | Nemotron 3 Ultra 550B A55B | NVIDIA | 81.4 | 1 | 2026-06-06 |
| 8 | Grok 4.20 0309 V2 | SpaceXAI | 81.2 | 3 | 2026-06-18 |
| 9 | Qwen3.7 Max | Alibaba | 80.5 | 3 | 2026-08-23 |
| 10 | Nemotron Cascade 2 30B A3B | NVIDIA | 80.4 | 2 | 2026-07-07 |
| 11 | Nova 2.0 Pro Preview | Amazon | 79.6 | 3 | 2026-05-02 |
| 12 | Qwen3.8 27B | Alibaba | 79.5 | 2 | 2026-08-23 |
| 13 | DeepSeek-V4-Flash | DeepSeek | 79.2 | 4 | 2026-08-24 |
| 14 | Qwen3.7 Plus Preview | Alibaba | 79.1 | 4 | 2026-08-23 |
| 15 | Qwen3.5 397B A17B | Alibaba | 78.8 | 4 | 2026-08-24 |
| 16 | Gemini 3 Flash Preview | 78 | 2 | 2026-05-02 | |
| 17 | GPT-5.2 Codex | OpenAI | 77.6 | 1 | 2026-05-02 |
| 18 | Gemini 3.1 Flash Lite Preview | 77.2 | 2 | 2026-07-07 | |
| 19 | Gemini 3.1 Pro | 77.1 | 3 | 2026-07-16 | |
| 20 | muse-glimmer-30b | Meta | 77 | 2 | 2026-08-23 |
| 21 | Qwen3.6 Max Preview | Alibaba | 76.6 | 1 | 2026-05-02 |
| 22 | DeepSeek-V4-Pro | DeepSeek | 76.5 | 4 | 2026-07-16 |
| 23 | Qwen3.5 27B | Alibaba | 76.5 | 3 | 2026-08-23 |
| 24 | Gemini 3.5 Flash | 76.3 | 4 | 2026-07-07 | |
| 25 | GLM-5.1 | Z.ai | 76.3 | 2 | 2026-05-02 |
| 26 | Qwen3.5 122B A10B | Alibaba | 76.1 | 3 | 2026-08-23 |
| 27 | Kimi K2.6 | Moonshot | 76 | 4 | 2026-07-16 |
| 28 | Muse Spark | Meta | 75.9 | 2 | 2026-07-07 |
| 29 | GPT-5.4 nano | OpenAI | 75.9 | 3 | 2026-06-01 |
| 30 | GPT-5.5 | OpenAI | 75.9 | 5 | 2026-06-01 |
| 31 | MiniMax M2.7 | MiniMax | 75.7 | 3 | 2026-08-24 |
| 32 | Gemma 4 31B | 75.6 | 2 | 2026-06-01 | |
| 33 | GPT-5 mini | OpenAI | 75.4 | 5 | 2026-07-07 |
| 34 | GPT-5.2 | OpenAI | 75.4 | 5 | 2026-08-24 |
| 35 | GPT-5.3 Codex | OpenAI | 75.4 | 1 | 2026-05-02 |
| 36 | Qwen3.6 Plus | Alibaba | 75.2 | 3 | 2026-08-23 |
| 37 | MAI-Code-1-Flash | Microsoft | 75 | 2 | 2026-08-23 |
| 38 | GPT-5 Codex | OpenAI | 74.1 | 1 | 2026-05-02 |
| 39 | Command A+ | Cohere | 74 | 3 | 2026-08-23 |
| 40 | GPT-5.4 | OpenAI | 73.9 | 3 | 2026-05-02 |
| 41 | Gemma 4 12B | 73.5 | 2 | 2026-06-14 | |
| 42 | GLM-5.2 Full Open Source | Z.ai | 73.3 | 2 | 2026-07-16 |
| 43 | GPT-5.4 mini | OpenAI | 73.3 | 3 | 2026-06-01 |
| 44 | GLM-5-Turbo | Z.ai | 73.2 | 2 | 2026-07-07 |
| 45 | GPT-5 | OpenAI | 73.1 | 7 | 2026-07-07 |
| 46 | GPT-5.1 | OpenAI | 72.9 | 2 | 2026-06-20 |
| 47 | GPT-5.6 Sol | OpenAI | 72.7 | 7 | 2026-08-12 |
| 48 | Nemotron 3 Super 120B A12B | NVIDIA | 72.6 | 2 | 2026-08-23 |
| 49 | Qwen3.5 35B A3B | Alibaba | 72.5 | 3 | 2026-08-23 |
| 50 | Gemma 4 26B A4B | 72.4 | 2 | 2026-05-06 | |
| 51 | GLM-5 | Z.ai | 72.3 | 2 | 2026-05-02 |
| 52 | MiniMax M2 | MiniMax | 72.3 | 4 | 2026-07-07 |
| 53 | MiMo V2 Flash | Xiaomi | 71.8 | 4 | 2026-07-07 |
| 54 | MiniMax M2.5 | MiniMax | 71.6 | 3 | 2026-08-24 |
| 55 | GPT-5.5 Instant | OpenAI | 71.5 | 2 | 2026-07-07 |
| 56 | o3 | OpenAI | 71.4 | 2 | 2026-07-07 |
| 57 | GPT-5.6 Terra | OpenAI | 71.2 | 5 | 2026-07-10 |
| 58 | Solar Pro 3 | Upstage | 71.2 | 2 | 2026-07-07 |
| 59 | Mercury 2 | Inception | 71 | 3 | 2026-08-23 |
| 60 | Nova 2.0 Lite | Amazon | 70.7 | 4 | 2026-06-01 |
| 61 | Gemini 3 Pro | 70.4 | 4 | 2026-08-24 | |
| 62 | o1 | OpenAI | 70.3 | 2 | 2026-07-07 |
| 63 | Kimi K2.5 | Moonshot | 70.2 | 3 | 2026-07-16 |
| 64 | MiniMax M2.1 | MiniMax | 70 | 5 | 2026-08-24 |
| 65 | GPT-5.1 Codex | OpenAI | 70 | 1 | 2026-05-02 |
| 66 | Apriel-v1.6-15B-Thinker | ServiceNow | 69.1 | 2 | 2026-07-07 |
| 67 | Qwen3.6 27B | Alibaba | 69.1 | 3 | 2026-08-14 |
| 68 | MAI-Thinking-1 | Microsoft | 69 | 1 | 2026-08-23 |
| 69 | GPT OSS 120B | OpenAI | 69 | 2 | 2026-05-02 |
| 70 | MiMo V2 Pro | Xiaomi | 68.8 | 2 | 2026-07-07 |
| 71 | o4-mini | OpenAI | 68.7 | 1 | 2026-05-02 |
| 72 | Kimi K2 (Reasoning) | Moonshot | 68.1 | 3 | 2026-07-07 |
| 73 | GLM-4.7 | Z.ai | 67.9 | 2 | 2026-06-01 |
| 74 | GPT-5.1 Codex mini | OpenAI | 67.9 | 1 | 2026-05-02 |
| 75 | GPT-5 nano | OpenAI | 67.6 | 5 | 2026-07-07 |
| 76 | MiMo V2 Omni | Xiaomi | 67.3 | 4 | 2026-07-07 |
| 77 | Step 3.7 Flash | StepFun | 67.3 | 2 | 2026-07-07 |
| 78 | MiMo V2.5 Pro | Xiaomi | 67.1 | 4 | 2026-08-24 |
| 79 | o3-mini | OpenAI | 67.1 | 1 | 2026-05-02 |
| 80 | Qwen3.5 9B | Alibaba | 66.7 | 5 | 2026-08-23 |
| 81 | Step 3.5 Flash | StepFun | 66.5 | 4 | 2026-07-07 |
| 82 | HyperNova 60B 2605 | Multiverse Computing | 66.5 | 2 | 2026-07-07 |
| 83 | Nex-N2-Pro | Nex AGI | 66.2 | 2 | 2026-07-07 |
| 84 | Nova 2.0 Omni Reasoning Medium | Amazon | 66.2 | 3 | 2026-06-01 |
| 85 | gpt-oss-20b | OpenAI | 65.1 | 3 | 2026-08-09 |
| 86 | K-EXAONE | LG AI | 64.7 | 2 | 2026-05-02 |
| 87 | Qwen3.6 35B A3B | Alibaba | 64.4 | 2 | 2026-06-01 |
| 88 | DeepSeek-V3.2-Speciale | DeepSeek | 63.9 | 2 | 2026-07-07 |
| 89 | Claude Fable 5 | Anthropic | 63.5 | 3 | 2026-08-12 |
| 90 | NVIDIA Nemotron 3 Nano 30B A3B | NVIDIA | 63.2 | 2 | 2026-06-01 |
| 91 | Kimi K2.7 Code | Moonshot | 63.1 | 2 | 2026-07-07 |
| 92 | Hy3-preview | Tencent | 63.1 | 2 | 2026-05-02 |
| 93 | K2 Think V2 | Moonshot | 62.8 | 1 | 2026-06-29 |
| 94 | Claude Opus 4.6 | Anthropic | 62.5 | 4 | 2026-08-24 |
| 95 | Claude Opus 4.8 | Anthropic | 62.2 | 2 | 2026-08-12 |
| 96 | Apriel-v1.5-15B-Thinker | ServiceNow | 61.7 | 2 | 2026-07-07 |
| 97 | GLM 5V Turbo | Z.ai | 61.1 | 1 | 2026-06-01 |
| 98 | DeepSeek-V3.2 | DeepSeek | 61 | 4 | 2026-06-06 |
| 99 | GLM-4.7-Flash | Z.ai | 60.8 | 2 | 2026-06-01 |
| 100 | Qwen3 Next 80B A3B | Alibaba | 60.7 | 1 | 2026-05-02 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.