APEX-Agents 1.1 Pass@1
Professional Work · 2026-09-08
Strict single-attempt success on APEX-Agents 1.1's 240-task cohort: a task succeeds only when every rubric criterion passes. The published estimate averages repeated attempts.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Argon | 82.4 |
| Claude Sonnet 5.5 | 75.5 |
| Claude Opus 5.5 | 73.5 |
| DeepSeek V4.1 Flash 1017 | 69.8 |
| Claude Fable 5.1 | 68.6 |
| GPT-6.1 Terra | 68.3 |
| Gemini 3.7 Flash | 67.8 |
| Kimi K3.1 Pro | 67.2 |
| Muse Wonder | 66.1 |
| Claude Opus 5 | 65.8 |
| Grok 4.6 | 65.3 |
| GPT-6 Astra | 64.7 |
| Gemini 3.8 Flash | 64.3 |
| Claude Fable 5 | 63.6 |
| GPT-6.1 Sol | 60.0 |
| Muse Spark 1.3 | 58.6 |
| GPT-5.6 Terra | 58.2 |
| GLM-5.3 | 56.6 |
| Grok 4.5 | 56.2 |
| Claude Haiku 5.5 | 56.1 |
| GPT-5.5 | 55.1 |
| Grok 4.7 | 54.6 |
| Sonnet 5 | 54.5 |
| MiMo-V2.6-Pro | 54.4 |
| GPT-6 Sol | 54.3 |
| MiniMax M3.1 Flash | 53.9 |
| GLM-5.3 Flash | 52.8 |
| GPT-5.4 | 52.4 |
| Mistral Large 4 Preview | 51.6 |
| GPT-5.6 Sol Pro | 51.5 |
| Mercury 3 | 50.2 |
| Opus 4.7 | 49.2 |
| Opus 4.8 | 48.9 |
| Beam 1.0 501B A23B | 48.8 |
| Kimi K3 | 47.8 |
| DeepSeek V4 Pro 0813 | 47.3 |
| Gemini 3.6 Flash | 46.9 |
| Opus 4.6 | 46.3 |
| Muse Spark 1.2 | 45.4 |
| GLM-5.2 | 45.2 |