Agents' Last Exam Unlicensed Overall Score
Agents · 2026-06-03
Average partial-credit rubric score across the 141-task unlicensed subset of Agents' Last Exam.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 56.0 |
| Kimi K3 | 53.0 |
| GPT-5.6 Terra | 52.3 |
| GPT-5.6 Luna | 52.0 |
| Claude Fable 5 | 50.2 |
| GPT-5.5 | 49.9 |
| Opus 4.8 | 46.9 |
| Opus 4.7 | 44.0 |
| GLM-5.2 | 42.4 |
| Doubao Seed 2.1 Pro | 42.2 |
| Composer 2.5 | 41.0 |
| GPT-5.4 | 39.9 |
| Gemini 3.1 Pro Preview | 34.1 |
| Qwen3.7-Max | 33.2 |
| DeepSeek-V4-Pro | 29.7 |
| GLM-5.1 | 29.7 |
| Qwen3.6 Plus (2026-04-02) | 25.7 |
| MiMo-V2.5 | 25.0 |
| Kimi K2.6 | 22.9 |
| Grok 4.3 | 21.6 |
| MiniMax M2.7 | 14.9 |