Qwen · Dense · Apache 2.0
Can I run Qwen2.5 14B locally?
The proven 14B before Qwen3. Still excellent and has more fine-tunes available today.
- Parameters
- 14.8B
- VRAM at Q4
- 11 GB
- Max context
- 131K
- Released
- 2024-09
Memory needed by quantisation
Total includes weights, an 8K-token KV cache and runtime overhead. Lower quants trade quality for size.
| Quant | Bits/weight | Weights | Total needed | Quality |
|---|---|---|---|---|
| FP16 | 16 | 28 GB | 31 GB | Full precision. Reference quality, 2× the size of 8-bit. |
| Q8_0 | 8.5 | 15 GB | 17 GB | Effectively lossless. The safe choice when it fits. |
| Q6_K | 6.56 | 11 GB | 14 GB | Near-lossless; quality loss is hard to measure. |
| Q5_K_M | 5.67 | 9.8 GB | 12 GB | Very good. A common sweet spot above Q4. |
| Q4_K_M | 4.83 | 8.3 GB | 11 GB | The default. Best size/quality trade-off for local use. |
| Q3_K_M | 3.91 | 6.7 GB | 9.3 GB | Noticeable degradation; useful to squeeze a size up. |
| Q2_K | 3.35 | 5.8 GB | 8.3 GB | Aggressive. Quality drops a lot — last resort to fit. |
Which hardware runs Qwen2.5 14B?
Best quantisation that fits each device at 8K context, with a rough speed estimate. Try your exact setup →
| Hardware | Memory | Fit | Best quant | Speed |
|---|---|---|---|---|
| RTX 3060 12 GB | 12 GB | Runs well | Q4 | ~29 tok/s |
| RTX 4060 Ti 16 GB | 16 GB | Runs well | Q6 | ~17 tok/s |
| RTX 5070 | 12 GB | Runs well | Q4 | ~54 tok/s |
| RTX 4070 Super | 12 GB | Runs well | Q4 | ~41 tok/s |
| Radeon RX 7900 XT | 20 GB | Runs well | Q8 | ~37 tok/s |
| RTX 5070 Ti | 16 GB | Runs well | Q6 | ~53 tok/s |
| RTX 4070 Ti Super | 16 GB | Runs well | Q6 | ~40 tok/s |
| RTX 3090 | 24 GB | Runs well | Q8 | ~43 tok/s |
| Radeon RX 7900 XTX | 24 GB | Runs well | Q8 | ~44 tok/s |
| RTX 5080 | 16 GB | Runs well | Q6 | ~57 tok/s |
| RTX 4080 Super | 16 GB | Runs well | Q6 | ~44 tok/s |
| RTX 4090 | 24 GB | Runs well | Q8 | ~46 tok/s |
| RTX 5090 | 32 GB | Runs well | FP16 | ~44 tok/s |
| RTX A6000 | 48 GB | Runs well | FP16 | ~19 tok/s |
| RTX 6000 Ada | 48 GB | Runs well | FP16 | ~23 tok/s |
| A100 80 GB | 80 GB | Runs well | FP16 | ~50 tok/s |
| H100 80 GB | 80 GB | Runs well | FP16 | ~81 tok/s |
| Mac · M1/M2/M3 (base), 16 GB | 16 GB | Runs well | Q4 | ~8.1 tok/s |
| Mac · M4 (base), 24 GB | 24 GB | Runs well | Q6 | ~7.1 tok/s |
| Mac · M4 Pro, 48 GB | 48 GB | Runs well | FP16 | ~6.6 tok/s |
| Mac · M1/M2/M3 Max, 32 GB | 32 GB | Runs well | Q8 | ~18 tok/s |
| Mac · M4 Max, 64 GB | 64 GB | Runs well | FP16 | ~13 tok/s |
| Mac · M1/M2/M3 Max, 64 GB | 64 GB | Runs well | FP16 | ~9.7 tok/s |
| Mac · M3/M4 Max, 128 GB | 128 GB | Runs well | FP16 | ~13 tok/s |
| Mac Studio · M1/M2 Ultra, 128 GB | 128 GB | Runs well | FP16 | ~19 tok/s |
| Mac Studio · M3 Ultra, 256 GB | 256 GB | Runs well | FP16 | ~20 tok/s |
| Mac Studio · M3 Ultra, 512 GB | 512 GB | Runs well | FP16 | ~20 tok/s |
| CPU only · 16 GB RAM | 16 GB | Runs well | Q5 | ~4.1 tok/s |
| CPU only · 32 GB RAM | 32 GB | Runs well | Q8 | ~3.2 tok/s |
| CPU only · 64 GB RAM | 64 GB | Runs well | FP16 | ~1.9 tok/s |
| CPU only · 128 GB RAM | 128 GB | Runs well | FP16 | ~2.2 tok/s |
| RTX 3080 10 GB | 10 GB | Runs (tight) | Q3 | ~76 tok/s |
| RTX 4060 Ti 8 GB | 8 GB | Won't fit | — | — |
| Mac · M1/M2/M3 (base), 8 GB | 8 GB | Won't fit | — | — |
| CPU only · 8 GB RAM | 8 GB | Won't fit | — | — |
GPUs that run Qwen2.5 14B well
The most affordable cards in our list that run it at a good quantisation.
RTX 3060 12 GB
12 GB · ~$279
Check price →RTX 4060 Ti 16 GB
16 GB · ~$449
Check price →RTX 5070
12 GB · ~$549
Check price →Hardware links are affiliate links — they don't change the recommendation.
What Qwen2.5 14B is good for
Related models
FAQ
How much VRAM does Qwen2.5 14B need?
At Q4_K_M, Qwen2.5 14B needs about 11 GB including a 8K-token context and overhead (8.3 GB for the weights alone). Higher quantisation needs more; see the table for every level.
What is the cheapest way to run Qwen2.5 14B?
The smallest device that runs it well in our list is the RTX 5070 (12 GB). Anything with at least that much memory should handle it at a usable quantisation.
Is Qwen2.5 14B good for general chat & assistance?
The proven 14B before Qwen3. Still excellent and has more fine-tunes available today.
Estimates — see how we compute these. Memory figures assume an 8K context; long-context use needs more.