Agents' Last Exam ALE-CLI Score
Agents · 2026-06-03
Average partial-credit rubric score on the 105-task ALE-CLI Linux-only subset of Agents' Last Exam.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 63.7 |
| GPT-6 Astra | 61.4 |
| GPT-6 Sol | 59.5 |
| Muse Spark 1.3 | 57.5 |
| Claude Opus 5 | 57.3 |
| MiniMax M3.1 Flash | 53.9 |
| GPT-5.6 Sol | 53.6 |
| Kimi K3 | 52.8 |
| GPT-5.6 Luna | 52.4 |
| GPT-6 Luna | 51.9 |
| Agnes 2.5 Pro Beta | 50.2 |
| GPT-5.6 Terra | 50.1 |
| GPT-5.5 | 49.2 |
| Claude Fable 5 | 48.5 |
| Opus 4.7 | 45.7 |
| Opus 4.8 | 44.6 |
| GLM-5.2 | 44.2 |
| Composer 2.5 | 42.2 |
| Doubao Seed 2.1 Pro | 42.2 |
| GPT-5.4 | 41.8 |
| Gemini 3.1 Pro Preview | 37.2 |
| Qwen3.7-Max (source unspecified) | 34.6 |
| Sonnet 4.6 | 32.0 |
| DeepSeek-V4-Pro | 30.8 |
| GLM-5.1 | 30.8 |
| Qwen3.6 Plus (2026-04-02) | 29.0 |
| MiMo-V2.5 | 26.5 |
| Grok 4.3 | 25.9 |
| Kimi K2.6 | 21.2 |
| MiniMax M2.7 | 14.6 |