ProgramBench (problem counts) - Almost
Code · 2026-05-05
The Almost split of ProgramBench (problem counts). This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 100 |
| Claude Sonnet 5.5 | 99.0 |
| GPT-6.1 Sol | 92.0 |
| Gemini 4 Argon | 82.0 |
| GPT-6 Sol | 65.0 |
| Claude Haiku 5.5 | 59.0 |
| GPT-5.6 Sol | 46.0 |
| Grok 4.7 | 43.0 |
| Grok 4.6 | 38.0 |
| Muse Spark 1.3 | 38.0 |
| Kimi K3 | 35.0 |
| MiMo-V2.6-Flash | 35.0 |
| DeepSeek V4.1 Flash | 33.0 |
| GLM-5.3 | 33.0 |
| Opus 4.8 | 31.0 |
| GPT-5.6 Terra | 30.0 |
| Hy4 Preview | 29.0 |
| Gemini 3.8 Flash | 28.0 |
| Sonnet 5 | 27.0 |
| DeepSeek V4 Pro 0813 | 27.0 |
| MiMo-V2.6-Pro | 25.0 |
| Qwen3.8-Max | 21.0 |
| Opus 4.7 | 19.0 |
| Gemini 3.7 Flash | 19.0 |
| GLM-5.2 | 19.0 |
| GPT-5.5 | 19.0 |
| GPT-5.6 Luna | 18.0 |
| DeepSeek-V4-Flash-0731 | 17.0 |
| GPT-6 Luna | 16.0 |
| Muse Wonder Full | 16.0 |
| GLM-5.3 Flash | 14.0 |
| GPT-5.4 | 14.0 |
| Grok 4.5 | 9.0 |
| GPT-5.4 Mini | 6.0 |
| Gemini 3.6 Flash | 5.0 |
| Muse Spark 1.1 | 5.0 |
| Muse Spark 1.2 | 5.0 |
| Sonnet 4.6 | 4.0 |
| Gemini 3.5 Flash | 3.0 |
| Gemini 3.1 Pro Preview | 2.0 |