Agents' Last Exam Unlicensed Last-Exam Score
Agents · 2026-06-03
Average partial-credit rubric score on the 30-task unlicensed Last-Exam subset of Agents' Last Exam.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3 | 23.1 |
| GPT-5.6 Sol | 20.7 |
| Claude Fable 5 | 16.4 |
| Opus 4.8 | 16.1 |
| GPT-5.6 Luna | 15.1 |
| GPT-5.5 | 15.0 |
| GPT-5.6 Terra | 15.0 |
| GLM-5.2 | 13.0 |
| Opus 4.7 | 12.8 |
| Composer 2.5 | 11.1 |
| GPT-5.4 | 10.3 |
| Doubao Seed 2.1 Pro | 9.1 |
| Qwen3.7-Max | 8.1 |
| GLM-5.1 | 6.2 |
| Qwen3.6 Plus (2026-04-02) | 4.6 |
| MiMo-V2.5 | 4.3 |
| Gemini 3.1 Pro Preview | 3.9 |
| DeepSeek-V4-Pro | 3.2 |
| Grok 4.3 | 3.0 |
| Kimi K2.6 | 2.1 |
| MiniMax M2.7 | 1.6 |