Advisor / VRAM tiers
Best local LLMs by VRAM
Pick your memory size to see which models fit, how much context you get, and how fast they run.
Memory tiers
| Tier | Fit fully | With offload | Largest that fits |
|---|---|---|---|
| 8 GB | 2 of 13 | 9 | Qwen3.5 9B Q4_K_M |
| 12 GB | 4 of 13 | 7 | Gemma 4 12B Q6_K |
| 16 GB | 5 of 13 | 6 | gpt-oss 20B Q8_0 |
| 24 GB | 11 of 13 | 0 | Qwen3.6 35B-A3B UD-Q4_K_M |
| 32 GB | 11 of 13 | 0 | Qwen3.6 35B-A3B UD-Q6_K |
| 48 GB | 11 of 13 | 1 | Qwen3.6 35B-A3B Q8_0 |
| 64 GB | 11 of 13 | 2 | Qwen3.6 35B-A3B Q8_0 |
| 96 GB | 13 of 13 | 0 | gpt-oss 120B Q8_0 |
| 128 GB | 13 of 13 | 0 | gpt-oss 120B Q8_0 |
| 192 GB | 13 of 13 | 0 | gpt-oss 120B Q8_0 |
| 256 GB | 13 of 13 | 0 | gpt-oss 120B Q8_0 |
| 512 GB | 13 of 13 | 0 | gpt-oss 120B Q8_0 |