AI Compute Radar · The card, seen throughEvery model on every card, as a table →
Qwen3 14B on RTX 4090
Qwen3 14B on RTX 4090 · Runs easily
Drag the picture to turn the card.
Memory ledger · what sits where
- File 8.4 GB
- Context 1.3 GB
- Runtime 1.1 GB
- Free 10.4 GB
- Margin 2.9 GB
- Weights ~8.4 GB (Q4_K_M), context ~1.3 GB at 8K tokens, runtime ~1.1 GB.
- Your 24 GB leaves ~21.1 GB usable after a 12% safety margin.
- Comfortable headroom — you can raise the context length or run other apps alongside it.
Length, height and thickness as the maker states them. Memory drawn as 12 modules of 2 GB; real boards differ in chip count.
File sizes measured from the published GGUF files, context memory computed from each model's attention shape, runtime assumed. No speed, watts or degrees. How the numbers are made →