Agents' Last Exam Full-Spectrum Pass Rate
Agents · 2026-06-03
Perfect-score pass rate on the 55-task full-spectrum subset, which provides broad coverage across more than 50 professional subdomains.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 30.0 |
| Kimi K3 | 29.1 |
| GPT-5.6 Luna | 27.3 |
| GPT-5.6 Terra | 27.3 |
| Claude Fable 5 | 25.5 |
| Doubao Seed 2.1 Pro | 24.1 |
| Opus 4.8 | 23.6 |
| GPT-5.5 | 23.6 |
| Opus 4.7 | 20.0 |
| GPT-5.4 | 19.4 |
| Composer 2.5 | 18.2 |
| GLM-5.2 | 18.2 |
| Gemini 3.1 Pro Preview | 12.7 |
| DeepSeek-V4-Pro | 10.9 |
| Qwen3.7-Max | 10.9 |
| GLM-5.1 | 9.1 |
| MiMo-V2.5 | 9.1 |
| Qwen3.6 Plus (2026-04-02) | 8.2 |
| Grok 4.3 | 7.3 |
| Kimi K2.6 | 6.4 |
| MiniMax M2.7 | 3.6 |