| Этап | Время | VRAM | RAM | Диск |
| Data prep (SFT) | ~2ч | — | ~20GB | ~2GB |
| SFT training (3 epochs) | ~8ч | ~20GB | ~8GB | ~5GB |
| SFT eval | ~1ч | ~16GB | ~4GB | — |
| GRPO reward setup | ~1ч | — | ~10GB | ~1GB |
| GRPO training (2 epochs) | ~18ч | ~22GB | ~60GB | ~10GB |
| Final eval + submission | ~2ч | ~16GB | ~4GB | ~100MB |