AHK-Eval Mid
Code · 2026-06-11
AHK-Eval Mid is the task-pass rate across the suite's 12 author-designated medium-difficulty tasks.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 100.0 |
| GPT-5.6 Sol Pro | 100.0 |
| GPT-5.5 | 100.0 |
| Grok 4.6 | 100.0 |
| GPT-5.6 Sol | 100.0 |
| GPT-5.6 Terra Pro | 100.0 |
| DeepSeek V4 Pro 0813 | 100.0 |
| GPT-5.6 Luna Pro | 91.7 |
| Claude Fable 5 | 91.7 |
| Kimi K3 | 91.7 |
| Gemini 3.7 Flash | 91.7 |
| GPT-5.6 Luna | 91.7 |
| Gemini 3.6 Flash | 91.7 |
| Gemini 3.1 Pro Preview | 91.7 |
| Kimi K2.6 | 91.7 |
| Opus 4.8 | 91.7 |
| Aion 3.0 | 91.7 |
| Claude Opus 5.5 | 83.3 |
| GPT-5.6 Terra | 83.3 |
| Grok 4.5 | 83.3 |
| Muse Spark 1.1 | 83.3 |
| Grok 4.3 | 83.3 |
| MiniMax M3 | 75.0 |
| Sonnet 4.6 | 75.0 |
| GPT-5.1 | 75.0 |
| Inkling | 75.0 |
| Hy3 | 75.0 |
| DeepSeek-V4-Pro | 75.0 |
| Gemini 3.5 Flash-Lite | 66.7 |
| GLM-5 | 58.3 |
| GLM-5.2 | 58.3 |
| Aion 3.0 Mini | 58.3 |
| LongCat-2.0 | 50.0 |
| Qwen3-Coder-480B-A35B-Instruct | 41.7 |
| Mistral Large 3 675B Instruct 2512 | 16.7 |
| Laguna XS 2.1 | 16.7 |