Can I run this LLM?
Can a Desktop CPU, DDR5 dual-channel (32GB) run Phi-4 14B?
Phi · MIT · estimated for a 32.0 GB machine at 4K context.
Yes, but it's tight.
Tight / slowBest fit: Phi-4 14B at Q2_K · 5.6 GB weights · 5–7 tok/s on a Desktop CPU, DDR5 dual-channel (32GB).
Arena Text Overall · style-controlled
1,256Q2_K untestedphi-4 · published configuration reference · Q2_K not tested
Arena Text Overall (style-controlled), exact published entry phi-4, 2026-10-08: reference rating 1,256.3. Published confidence bounds 1,251.7–1,260.9. 24,126 votes. Published phi-4 configuration reference. Reasoning mode is not disclosed by the Arena dataset. Tested precision, runtime and sampling settings are not disclosed in this dataset. Selected Q2_K has not been measured by this Arena entry. No quantization quality adjustment is applied.
Snapshot published 2026-10-08 · snapshot imported 2026-10-11 UTC · validated last-good data · manual refresh
Arena leaderboard dataset · Arena / lmarena-ai, Arena Leaderboard Dataset (CC BY 4.0). Selected Text Overall style-controlled rows; field names adapted, ratings unchanged. · dataset CC BY 4.0 · revision e16f90232986
- Total memory at 4K context
- 7.2 GB
- Weights + KV cache + runtime overhead
- Estimated generation
- 5–7 tok/s
- Estimated first token
- ~3m
Every quantization on this machine
On small screens, each quant lists every metric below its name. On wider screens, scroll sideways for total memory, speed and fit. Keyboard: focus the table, then use the arrow keys. Quant names stay visible.
| Quant | Weights | Total memory | Speed | Fit |
|---|---|---|---|---|
| Q2_K | 5.6 GB | 7.2 GB | 5–7 tok/s | Tight / slow |
| Q3_K_M | 7.2 GB | 8.8 GB | 4–6 tok/s | Tight / slow |
| Q4_K_M | 8.9 GB | 10.5 GB | 3–5 tok/s | Tight / slow |
| Q5_K_M | 10.4 GB | 12.0 GB | 3–4 tok/s | Tight / slow |
| Q6_K | 12.0 GB | 13.7 GB | 2–4 tok/s | Tight / slow |
| Q8_0 | 15.6 GB | 17.2 GB | 2–3 tok/s | Tight / slow |
| FP16 | 29.3 GB | 31.0 GB | — not resident | Won't fit |
Ran Phi-4 14B on a Desktop CPU, DDR5 dual-channel (32GB)? Add your real numbers →