IFBench (prompt loose)
IFBench (prompt loose) - instruction following benchmark, loose prompt evaluation.
Category
Instruction Following
Max Score
100.0
Score Type
percent
Active
Yes
Model Scores
Scores for IFBench (prompt loose)
| Model | Score | Date | Verified |
|---|---|---|---|
| NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 |
98.46%
|
04.06.2026 | Verified |
| MiniMax-M2.7 |
62.05%
|
04.06.2026 | Unverified |
| GLM-5.1 |
72.31%
|
04.06.2026 | Unverified |
| Kimi K2.6 |
57.44%
|
04.06.2026 | Unverified |
| Qwen3.5-397B-A17B |
80.51%
|
04.06.2026 | Unverified |
| DeepSeek-V4-Pro |
85.13%
|
04.06.2026 | Unverified |
| DeepSeek-V4-Flash |
100.00%
|
04.06.2026 | Unverified |
| Qwen3.8-27B |
87.18%
|
18.08.2026 | Verified |
| Qwen3.6-27B |
33.85%
|
18.08.2026 | Unverified |
| Qwen3.7-Plus |
85.13%
|
18.08.2026 | Unverified |
| Muse Glimmer-30B |
74.36%
|
18.08.2026 | Unverified |
| Opus4.6 Max |
62.50
|
18.08.2026 | Unverified |