AI BENCHY Category Failures
Instructions following: Wrong answer
Instructions following
Wrong answer
See which AI models are most likely to hit Wrong answer on Instructions following, so you can spot weak points faster.
Failure Reasons
| Rank | Model | Company | Wrong answer Count | Category Score | Tests Correct | Response Time (avg) |
|---|---|---|---|---|---|---|
| #86 | GPT-5.4 Mini none | OpenAI | 1 | 6.3 | 1/2 | 728ms |
| #88 | Nemotron 3 Super none | NVIDIA | 1 | 4.9 | 0/2 | 1.50s |
| #89 | GPT-4o-mini none | OpenAI | 1 | 4.8 | 0/2 | 1.27s |
| #90 | Qwen3.5-9B none | Qwen | 1 | 6.5 | 1/2 | 514ms |
| #91 | Mercury 2 none | Inception | 1 | 6.5 | 1/2 | 551ms |
| #92 | Qwen3 Coder Next medium | Qwen | 1 | 4.8 | 0/2 | 7.34s |
| #93 | GLM 4.7 Flash medium | Z.ai | 1 | 6.2 | 1/2 | 2.97s |
| #94 | MiMo-V2-Flash none | Xiaomi | 1 | 6.5 | 1/2 | 857ms |
| #96 | GPT-5.4 Nano none | OpenAI | 1 | 5.0 | 0/2 | 787ms |