| Model Size | VRAM (FP16) | VRAM (4-bit) | Best Use Cases

| – | — | | – |

| 1–3B | 4–6 GB | \~2 GB | Basic chat, text classification, autocomplete |
| 7–8B | 14–16 GB | \~6–8 GB | General-purpose assistants, summarization, coding |
| 13–14B | 26–28 GB | \~12–16 GB | Stronger reasoning, instruction following |
| 70B+ | 140+ GB | \~35–40 GB | Enterprise-level reasoning, SOTA applications |