Microsoft's 14B that reasons above its weight, especially at math. 16K context.
reasoninggeneralcoding · sizes: 14B · MIT
| machine | best fit | verdict | speed |
|---|---|---|---|
| Apple M1 · 8GB | 14B Q3_K_M | ❌ Won't fit | — |
| Apple M1 Pro · 16GB | 14B Q5_K_M | 🐢 CPU-only (slow) | 1–2 tok/s |
| Apple M2 · 16GB | 14B Q5_K_M | 🐢 CPU-only (slow) | 0.7–1 tok/s |
| Apple M3 Pro · 18GB | 14B Q3_K_M | ✅ Runs (tight) | 11–19 tok/s |
| Apple M2 Pro · 32GB | 14B Q8_0 | ✅ Runs (tight) | 7–11 tok/s |
| Apple M4 Pro · 48GB | 14B Q8_0 | ✅ Runs comfortably | 9–15 tok/s |
| Apple M3 Max · 36GB | 14B Q8_0 | ✅ Runs comfortably | 12–20 tok/s |
| Apple M2 Max · 64GB | 14B Q8_0 | ✅ Runs comfortably | 12–20 tok/s |
| Apple M4 Max · 128GB | 14B Q8_0 | ✅ Runs comfortably | 15–25 tok/s |
| Apple M2 Ultra · 192GB | 14B Q8_0 | ✅ Runs comfortably | 20–35 tok/s |
| NVIDIA GeForce RTX 3060 · 12GB VRAM | 14B Q4_K_M | ✅ Runs (tight) | 16–25 tok/s |
| NVIDIA GeForce RTX 4070 · 12GB VRAM | 14B Q4_K_M | ✅ Runs (tight) | 25–40 tok/s |
| NVIDIA GeForce RTX 4090 · 24GB VRAM | 14B Q8_0 | ✅ Runs comfortably | 30–45 tok/s |
| NVIDIA GeForce RTX 5080 · 16GB VRAM | 14B Q5_K_M | ✅ Runs (tight) | 35–60 tok/s |
| NVIDIA GeForce RTX 5090 · 32GB VRAM | 14B Q8_0 | ✅ Runs comfortably | 50–85 tok/s |
| AMD Radeon RX 7900 XTX · 24GB VRAM | 14B Q8_0 | ✅ Runs comfortably | 25–40 tok/s |
Shopping for a machine? What hardware do I need to run Phi-4? →
Check your exact rig: open the checker →