Benchmark
IFEval
Instruction following benchmarks measure how closely models follow instructions. Scores are percentages. Higher is better.
- Models measured
- 209
- Scores recorded
- 219
- Sources
- 3
- IQ weight
- 0.84
Score distribution
- 57.9%
- Mean
- 59.7%
- Median
- 26.2%
- Std dev
- 20.6%
- 10th percentile
- 89.8%
- 90th percentile
- 0%
- Min
This benchmark has a weight of 0.84 for chat models, measured over 209 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Gemma 4 31B ITGemma 4 Technical Report · Table 5 · thinking mode unless noted· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 98.9% |
| 2 | Gemma 4 26B A4BGemma 4 Technical Report · Table 5 · thinking mode unless noted· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 98.5% |
| 3 | Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 95% |
| 4 | Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 94.6% |
| 5 | Qwen3.7 PlusQwen3.7-Plus release · recommended reasoning-effort=xhigh system prompt· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 94.6% |
| 6 | Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 94.3% |
| 7 | Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise · strict prompt· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 94.3% |
| 8 | Qwen3.7 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 94.3% |
| 9 | Qwen3.7 MaxQwen3.7-Max release · recommended reasoning-effort=xhigh system prompt· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 94.3% |
| 10 | o3 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 93.9% |
| 11 | Qwen3 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 93.4% |
| 12 | Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 93.4% |
| 13 | Claude Sonnet 3.7· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 93.2% |
| 14 | Qwen3.5 397B A17B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 92.6% |
| 15 | Qwen3.5 397B A17BQwen3.5 release/model card · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 92.6% |
| 16 | Llama 3.3 70B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 92.1% |
| 17 | Nova Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 92.1% |
| 18 | Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 91.9% |
| 19 | Qwen3.5 9B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 91.5% |
| 20 | Gemma 3 27B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 90.4% |
| 21 | Nemotron Nano 9b V2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 90.3% |
| 22 | Gemma 3 4B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 90.2% |
| 23 | Kimi K2 Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 89.8% |
| 24 | Qwen3.5 4B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 89.8% |
| 25 | Nova Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 89.7% |
| 26 | Qwen3.5 Omni PlusQwen3.5-Omni release · no-thinking setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 89.7% |
| 27 | LongCat Flash Chat· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 89.7% |
| 28 | Llama 3.1 Nemotron Ultra 253B v1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 89.5% |
| 29 | Gemma 3 12B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 88.9% |
| 30 | Qwen3 Next 80B A3Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 88.9% |
| 31 | Qwen3 235B A22B Instruct 2507· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 88.7% |
| 32 | Llama 3.1 405B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 88.6% |
| 33 | GPT-4.5 Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 88.2% |
| 34 | Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 88.2% |
| 35 | Qwen 3 235B A22B Thinking 2507· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.8% |
| 36 | Qwen3 VL 235B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 87.8% |
| 37 | Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 87.8% |
| 38 | Qwen3 Next 80B A3B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.6% |
| 39 | Llama 3.1 70B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 87.5% |
| 40 | Llama 3.1 70B Instructofficial instruction-tuned table· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.5% |
| 41 | GPT 4.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.4% |
| 42 | Nova Micro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.2% |
| 43 | Qwen2.5 72B Instruct· Hugging Face Open LLM Leaderboard dataValues from 2 sources
| 86.4% |
| 44 | DeepSeek V3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 86.1% |
| 45 | qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 85.8% |
| 46 | Qwen3 14B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 85.4% |
| 47 | Qwen3.5 Omni FlashQwen3.5-Omni release · no-thinking setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.2% |
| 48 | Phi 4 Reasoning Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.9% |
| 49 | Sarvam 105B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.8% |
| 50 | Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 84.7% |
Showing 1–50 of 219 results.
Data credits: Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.