From the RTX 3070 8 GB to the RTX 5060 Ti 16 GB: what changes for local AI
+8 GB of VRAM (8 → 16 GB). Of 49 models, 17 newly run well, 26 get a better file or stop offloading, and 6 stay the same.
17newly run well
26better file
6no change
+8GB more VRAM
01Newly runs well
| Model | On the RTX 3070 8 GB | On the RTX 5060 Ti 16 GB | Needed |
|---|---|---|---|
| FLUX.1 [dev] | Tight Q3_K_S | Runs well FP8 | 14.2 GB |
| FLUX.1 [schnell] | Tight Q3_K_S | Runs well FP8 | 14.2 GB |
| FLUX.1 Kontext [dev] | Tight Q3_K_M | Runs well FP8 | 14.5 GB |
| FLUX.1 Krea [dev] | Tight Q3_K_M | Runs well FP8 | 14.2 GB |
| FLUX.1 Fill [dev] | Tight Q3_K_S | Runs well Q8_0 | 15.3 GB |
| FLUX.2 [klein] 9B | Tight Q3_K_M | Runs well FP8 | 11.7 GB |
| Krea 2 (Turbo) | Tight Q2_K | Runs well FP8 | 15.7 GB |
| Qwen-Image 2.1 | Runs Q5_K_M | Runs well Q8_0 | 9.9 GB |
| Z-Image Turbo | Runs Q6_K | Runs well 16-bit | 14.3 GB |
| Z-Image (base) | Runs Q5_K_M | Runs well 16-bit | 14.3 GB |
| Boogu-Image (Turbo) | Offload only Q4_1 | Runs well FP8 | 12.9 GB |
| ERNIE-Image (Turbo) | Runs Q4_K_M | Runs well Q8_0 | 11.0 GB |
| HiDream-O1-Image | Offload only FP8 | Runs well FP8 | 10.9 GB |
| Stable Diffusion 3.5 Large | Runs Q4_1 | Runs well Q8_0 | 11.1 GB |
| Chroma1-HD | Runs Q4_K_M | Runs well FP8 | 11.5 GB |
| Wan 2.2 TI2V 5B | Runs Q5_K_M | Runs well 16-bit | 13.8 GB |
| HunyuanVideo 1.5 | Offload only Q4_K_M | Runs well FP8 | 12.6 GB |
“Runs well” means a 16-bit or 8-bit file fits entirely in VRAM.
02A better file, or less offloading
| Model | On the RTX 3070 8 GB | On the RTX 5060 Ti 16 GB | Needed |
|---|---|---|---|
| Qwen-Image | Offload only Q2_K | Runs Q4_K_M | 15.9 GB |
| Qwen-Image-Edit (2511) | Offload only Q4_K_M | Tight Q3_K_M | 12.9 GB |
| Ideogram 4 | Offload only Q4_1 | Runs Q4_1 | 14.7 GB |
| HiDream-I1 (Full) | Offload only Q2_K | Runs Q5_K_M | 15.8 GB |
| HiDream-I1 (Dev) | Offload only Q2_K | Runs Q5_K_M | 15.8 GB |
| HunyuanImage 2.1 | Offload only Q4_K_M | Runs Q4_K_M | 14.6 GB |
| Wan 2.1 T2V 14B | Offload only Q4_K_M | Runs Q5_K_M | 15.6 GB |
| Wan 2.1 I2V 14B 480P | Offload only Q4_K_M | Runs Q4_K_M | 15.6 GB |
| Wan 2.1 I2V 14B 720P | Offload only Q4_K_M | Tight Q3_K_M | 15.4 GB |
| Wan 2.1 VACE 14B | Offload only Q4_K_M | Tight Q3_K_S | 12.6 GB |
| Wan 2.2 T2V A14B | Offload only Q2_K | Runs Q5_K_M | 15.1 GB |
| Wan 2.2 I2V A14B | Offload only Q2_K | Runs Q5_K_M | 15.1 GB |
| Wan 2.2 Animate 14B | Offload only Q4_K_M | Tight Q3_K_M | 13.9 GB |
| Wan Animate 2 (14B) | Offload only Q4_K_M | Tight Q3_K_M | 13.9 GB |
| Wan 2.2 S2V 14B | Offload only Q4_K_M | Tight Q2_K | 14.3 GB |
| SCAIL-2 (character animation) | Offload only Q4_K_M | Tight Q3_K_M | 14.4 GB |
| HunyuanVideo (13B, original) | Offload only Q4_K_M | Runs Q6_K | 15.3 GB |
| LTX-Video 13B (0.9.8) | Offload only Q2_K | Runs Q6_K | 15.2 GB |
| LTX-2 (19B) | Offload only Q4_K_M | Tight Q3_K_M | 14.9 GB |
| LTX-2.3 (22B) | Offload only Q4_K_M | Tight Q3_K_M | 15.6 GB |
| LTX-2.5 (22B) | Offload only Q4_K_M | Tight Q2_K | 13.6 GB |
| MiniMax H3 Pruned | Offload only Q4_K_M | Tight Q3_K_M | 14.7 GB |
| FLUX.2 [dev] | Not practical Q2_K | Offload only Q2_K | 16.2 GB |
| FLUX.2 [klein] 4B | Runs well Q8_0 | Runs well 16-bit | 9.6 GB |
| Mage-Flow (Microsoft) | Runs well INT8 | Runs well 16-bit | 10.0 GB |
| MiniMax H3 (33B) | Not practical Q3_K_M | Offload only Q4_K_M | 25.7 GB |
03No change in what fits
Lumina 2.0 Runs wellSD 3.5 Medium Runs wellSDXL Runs wellIllustrious / Pony Runs wellSD 1.5 Runs wellWan 2.1 1.3B Runs well
Same verdict and same file on both cards. Speed can still differ.
04Speed and features
Memory bandwidth goes from 448 to 448 GB/s (×1.00). For models that fit on both cards, bandwidth and compute decide the speed; this ratio is a rough first guide, not a benchmark. The RTX 5060 Ti 16 GB computes FP8 natively, so FP8 files run faster there than on the RTX 3070 8 GB, which only uses them to save memory. It also adds FP4 (NVFP4) hardware, for models that ship NVFP4 files.