Dataset update
The public dataset now covers 24 model entries and 9 hardware presets. This update adds Qwen3.6 27B and a 32GB Blackwell workstation preset for RTX 5090-class local AI planning.
Qwen3.6 27B now appears in the calculator, model index, Ollama reference, GPU tables, and agent-readable JSON. The new 32GB preset uses RTX 5090 as the example GPU and exposes about 30.5GB of clean planning VRAM after the standard system reserve.
For local AI users, 32GB is a useful bridge between 24GB consumer workstations and 48GB professional cards. It fits 27B-32B Q4 models more comfortably, but 70B+ models and very long context runs still deserve cloud/API fallback checks.