Microsoft on NVIDIA Ada
Can I run Phi-4 14B on an RTX 4070 Ti Super?
Yes. Phi-4 14B at Q4_K_M uses 10.8 GB of the 14.2 GB available on a RTX 4070 Ti Super, and runs at about 43.5 tokens per second. You can push the context to 16K.
Runs comfortably
10.8 GB of 14.2 GB · 76%014 GB
Weights
8.4 GB
KV cache
1.6 GB
Runtime overhead
0.8 GB
Phi-4 14B at Q4_K_M leaves 3.5 GB spare on a RTX 4070 Ti Super. There is room to raise the context length or move up a quantisation level.
Generation43.5tok/s
Prompt processing1257tok/s
Max context16Ktokens
KV per 1K tokens0GB
Every quantisation of Phi-4 14B on a RTX 4070 Ti Super
Highlighted row is the highest quality that still fits at 8K context.
| Quantisation | bpw | Weights | Total | Verdict | Max ctx | tok/s |
|---|---|---|---|---|---|---|
| FP16 / BF16 | 16.00 | 27.4 GB | 29.8 GB | Won't fit | — | 2.21 |
| INT8 / W8A8 | 8.50 | 14.3 GB | 16.7 GB | Won't fit | — | 9.98 |
| Q8_0 (GGUF) | 8.50 | 14.3 GB | 16.7 GB | Won't fit | — | 9.98 |
| FP8 (E4M3) | 8.00 | 13.7 GB | 16.1 GB | Won't fit | — | 11.4 |
| Q6_K | 6.56 | 11.2 GB | 13.6 GB | Fits, but tight | 11K | 33.3 |
| Q5_K_M | 5.67 | 9.7 GB | 12.1 GB | Runs comfortably | 16K | 38.1 |
| Q5_K_S | 5.52 | 9.4 GB | 11.8 GB | Runs comfortably | 16K | 39.0 |
| AWQ 4-bit | 4.25 | 8.7 GB | 11.1 GB | Runs comfortably | 16K | 42.2 |
| GPTQ 4-bit | 4.25 | 8.7 GB | 11.1 GB | Runs comfortably | 16K | 42.2 |
| MXFP4 | 4.25 | 8.7 GB | 11.1 GB | Runs comfortably | 16K | 42.2 |
| Q4_K_M | 4.85 | 8.4 GB | 10.8 GB | Runs comfortably | 16K | 43.5 |
| Q4_K_S | 4.58 | 7.9 GB | 10.3 GB | Runs comfortably | 16K | 45.7 |
| Q4_0 | 4.55 | 7.9 GB | 10.3 GB | Runs comfortably | 16K | 45.9 |
| IQ4_XS | 4.25 | 7.4 GB | 9.8 GB | Runs comfortably | 16K | 48.5 |
| Q3_K_M | 3.91 | 6.9 GB | 9.3 GB | Runs comfortably | 16K | 51.9 |
| IQ3_M | 3.70 | 6.5 GB | 8.9 GB | Runs comfortably | 16K | 54.3 |
| IQ3_XXS | 3.06 | 5.5 GB | 7.9 GB | Runs comfortably | 16K | 62.9 |
| Q2_K | 2.63 | 4.8 GB | 7.2 GB | Runs comfortably | 16K | 70.4 |
| IQ2_XXS | 2.06 | 3.9 GB | 6.3 GB | Runs comfortably | 16K | 83.7 |
| IQ1_M | 1.75 | 3.4 GB | 5.8 GB | Runs comfortably | 16K | 93.3 |