APEX-v1-extended
Professional Work · 2025-12-10
Mercor APEX-v1-extended is a held-out professional-work benchmark that extends APEX with expert-evaluated tasks. Scores report percentage task performance.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.4 | 67.2 |
| Opus 4.6 | 65.7 |
| GPT-5.2-Codex | 65.3 |
| Gemini 3.1 Pro Preview | 65.1 |
| GPT-5.2 | 64.7 |
| GPT-5.3-Codex | 64.6 |
| GPT-5 | 64.5 |
| GPT-5-Codex | 62.8 |
| GPT-5.1-Codex | 61.9 |
| Gemini 3 Flash Preview | 60.9 |
| Gemini 3 Pro Preview | 60.9 |
| o3 | 60.5 |
| Grok 4.1 Fast | 57.5 |
| Opus 4.5 | 57.3 |
| Grok 4 | 57.1 |
| Sonnet 4.6 | 54.4 |
| GLM-4.7 | 51.7 |
| GLM-5 | 49.0 |