Best local AI models for 8GB VRAM
Ranked by what actually fits at 32K context, computed from real file bytes.
A 8GB card gives you about 7.44 GiB to work with after driver overhead. 8 indexed models fit at 32K context — the largest being Wan2.2-Animate-14B at 17.3B parameters in Q2_K.
From the file· fit from summed bytesFrom the file· KV per layer
Fits in 8GB at 32K context
largest quantization that fits, per model
| Model | Modality | Best quant | Params○ | Total◐ | Headroom◐ |
|---|---|---|---|---|---|
| Wan2.2-Animate-14B | video generation | Q2_K | 17.3B | 7.20 GiB | 0.24 GiB |
| Bernini-R | video generation | Q3_K_S | 14.3B | 6.91 GiB | 0.53 GiB |
| Wan2.2-Distill-Models | video generation | Q3_K_S | 14.3B | 6.91 GiB | 0.53 GiB |
| Wan2.2-TI2V-5B | video generation | Q8_0 | 5.0B | 5.87 GiB | 1.57 GiB |
| Wan2.1-T2V-14B | video generation | Q3_K_S | 14.3B | 7.34 GiB | 0.10 GiB |
| HunyuanVideo-1.5 | video generation | Q6_K | 8.3B | 7.39 GiB | 0.05 GiB |
| Wan2.2-TI2V-5B-Turbo | video generation | Q8_0 | 5.0B | 5.87 GiB | 1.57 GiB |
| SkyReels-V2-DF-14B-540P | video generation | Q3_K_S | 14.3B | 6.91 GiB | 0.53 GiB |
This page models a generic 8GB accelerator, so it answers what fits rather than how fast it runs. For tokens per second you need a specific card — pick one from hardware, where bandwidth is known.