IFBench (Artificial Analysis)
Chat & Writing · 2025-07-02
IFBench evaluates precise instruction following on the 294-question Artificial Analysis implementation of AllenAI's 58 diverse, verifiable out-of-domain constraints. This evaluator-specific component measures whether models generalize to new output rules and formatting demands; higher scores mean more constraints satisfied.
Top models (higher is better)
| Model | Score |
|---|---|
| Inkling-Small Preview | 83.4 |
| Grok 4.3 | 83.3 |
| Grok 4.20 | 82.9 |
| MiniMax M3 | 82.9 |
| Inkling-Small | 82.2 |
| Nemotron 3 Ultra 550B A55B | 81.4 |
| Qwen3.7-Max | 80.5 |
| Nemotron Cascade 2 30B A3B | 80.4 |
| MiMo-V2.5-Pro | 79.9 |
| Inkling | 79.8 |
| Nova 2 Pro (Preview) | 79.6 |
| DeepSeek-V4-Flash | 79.2 |
| Qwen3.5 397B A17B | 78.8 |
| Gemini 3.5 Flash-Lite | 78.6 |
| Gemini 3 Flash Preview | 78.0 |
| Qwen3.7-Plus | 78.0 |
| GPT-5.2-Codex | 77.6 |
| Gemini 3.1 Flash-Lite Preview | 77.2 |
| Gemini 3.1 Pro Preview | 77.1 |
| Qwen3.6-Max-Preview | 76.6 |
| DeepSeek-V4-Pro | 76.5 |
| Gemini 3.5 Flash | 76.3 |
| GLM-5.1 | 76.3 |
| Kimi K2.6 | 76.0 |
| GPT-5.4 Nano | 75.9 |
| Muse Spark | 75.9 |
| GPT-5.5 | 75.9 |
| MiniMax M2.7 | 75.7 |
| Qwen3.5 122B A10B | 75.7 |
| Gemma 4 31B IT | 75.6 |
| Qwen3.5-27B | 75.6 |
| GPT-5.2 | 75.4 |
| GPT-5 Mini | 75.4 |
| GPT-5.3-Codex | 75.4 |
| Qwen3.6 Plus (2026-04-02) | 75.2 |
| GPT-5-Codex | 74.1 |
| Command A+ (May 2026, BF16) | 73.9 |
| GPT-5.4 | 73.9 |
| Gemma 4 12B IT | 73.5 |
| GLM-5.2 | 73.3 |