Agents' Last Exam Unlicensed Near-Term Score
Agents · 2026-06-03
Average partial-credit rubric score on the 67-task unlicensed near-term subset of Agents' Last Exam.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 78.8 |
| GPT-5.6 Luna | 74.1 |
| GPT-5.6 Terra | 73.9 |
| GPT-5.5 | 72.4 |
| Kimi K3 | 71.6 |
| Claude Fable 5 | 71.1 |
| Opus 4.8 | 64.0 |
| Opus 4.7 | 61.2 |
| Composer 2.5 | 61.1 |
| GLM-5.2 | 59.1 |
| Doubao Seed 2.1 Pro | 58.9 |
| GPT-5.4 | 57.8 |
| Gemini 3.1 Pro Preview | 53.5 |
| Qwen3.7-Max | 46.9 |
| GLM-5.1 | 45.6 |
| DeepSeek-V4-Pro | 43.8 |
| Qwen3.6 Plus (2026-04-02) | 35.8 |
| Kimi K2.6 | 35.1 |
| MiMo-V2.5 | 35.1 |
| Grok 4.3 | 30.4 |
| MiniMax M2.7 | 24.5 |