← Back to Test Cases
According to ICH Q7, which of the following best describes the requirements for impurity profiles of APIs?
Knowledge QAIch Q7hard
Cross-Model Comparison
| Model | Score | Latency | Tokens In | Tokens Out |
|---|---|---|---|---|
| Gemini 3.1 Flash-Lite | 100.0% | 1.8s | 104 | 214 |
| GPT-5.4 mini | 100.0% | 772ms | 112 | 66 |
| GPT-5.4 nano | 100.0% | 641ms | 112 | 44 |
| Claude Haiku 4.5 | 100.0% | 4.0s | 128 | 358 |
| Claude Sonnet 4.6 | 100.0% | 7.9s | 128 | 316 |
| Claude Opus 4.6 | 100.0% | 11.6s | 128 | 420 |
| Llama 4 Scout | 100.0% | 1.1s | 114 | 410 |
| Llama 4 Maverick | 100.0% | 3.4s | 183 | 394 |
| Llama 3.3 70B Instruct | 100.0% | 10.3s | 120 | 195 |
| Mistral Small 2603 | 100.0% | 2.0s | 119 | 308 |
| Gemini 3 Flash | 100.0% | 1.9s | 103 | 220 |
| GPT-5.4 | 100.0% | 3.1s | 112 | 205 |
| DeepSeek-V3.2 | 100.0% | 4.8s | 113 | 126 |
| DeepSeek-R1 | 100.0% | 22.4s | 113 | 852 |
| DeepSeek-R1-Distill-Qwen-32B | 100.0% | 26.4s | 112 | 595 |
| Mistral Large 3 675B | 100.0% | 6.2s | 107 | 355 |
| Gemini 3.1 Pro | 100.0% | 13.7s | 103 | 848 |
| Qwen3.5-397B-A17B | 0.0% | 44.1s | 120 | 1,288 |
| Qwen3.5-35B-A3B | 0.0% | 6.7s | 120 | 1,043 |
Tags
impurity_managementimpurity_profilequality_control