What GPU do you need to run Qwen 2.5 Coder 14B?
14.7B params 128K max context
The smallest GPU memory that runs Qwen 2.5 Coder 14B well is 12 GB (Q4_K_M, e.g. GeForce RTX 5070). 86 of 132 GPUs run it well and 4 more are a tight fit.
MMLU 74 SWE-bench Verified 22.6
VRAM needed per quantization
| Quantization | Weights | 16K context | 32K context | Full 128K |
|---|---|---|---|---|
| Q4_K_M | 9.0 GB | 11.9 GB | 14.8 GB | 32.4 GB |
| Q8_0 | 15.7 GB | 18.6 GB | 21.6 GB | 39.1 GB |
Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.
Cheapest GPUs that can run Qwen 2.5 Coder 14B
Lowest launch MSRP among GPUs that run its smallest quantization (Q4_K_M) well, with at least 16K context.
As an Amazon Associate, I earn from qualifying purchases.
GPUs that can run Qwen 2.5 Coder 14B
Best quantization that fits on each GPU, with the context and speed it gets. For chips with several memory sizes, the smallest that works is shown.
NVIDIA
- RTX 3080 10 GB Q4_K_M 5K context · ~76 tok/s tight fit
- GTX 1080 Ti 11 GB Q4_K_M 10K context · ~48 tok/s tight fit
- RTX 2080 Ti 11 GB Q4_K_M 10K context · ~62 tok/s tight fit
- RTX 2060 12GB 12 GB Q4_K_M 16K context · ~34 tok/s
- RTX 3060 12 GB Q4_K_M 16K context · ~36 tok/s
- RTX 3080 12GB 12 GB Q4_K_M 16K context · ~91 tok/s
- RTX 3080 Ti 12 GB Q4_K_M 16K context · ~91 tok/s
- RTX 4070 12 GB Q4_K_M 16K context · ~50 tok/s
- RTX 4070 SUPER 12 GB Q4_K_M 16K context · ~50 tok/s
- RTX 4070 Ti 12 GB Q4_K_M 16K context · ~50 tok/s
- RTX 5070 12 GB Q4_K_M 16K context · ~67 tok/s
- RTX A2000 12GB 12 GB Q4_K_M 16K context · ~29 tok/s
- RTX 4060 Ti 16GB 16 GB Q4_K_M 38K context · ~29 tok/s
- RTX 4070 Ti SUPER 16 GB Q4_K_M 38K context · ~67 tok/s
- RTX 4080 16 GB Q4_K_M 38K context · ~72 tok/s
- RTX 4080 SUPER 16 GB Q4_K_M 38K context · ~74 tok/s
- RTX 5060 Ti 16GB 16 GB Q4_K_M 38K context · ~45 tok/s
- RTX 5070 Ti 16 GB Q4_K_M 38K context · ~90 tok/s
- RTX 5080 16 GB Q4_K_M 38K context · ~96 tok/s
- RTX 2000 Ada 16 GB Q4_K_M 38K context · ~22 tok/s
- RTX A4000 16 GB Q4_K_M 38K context · ~45 tok/s
- RTX PRO 2000 Blackwell 16 GB Q4_K_M 38K context · ~29 tok/s
- Tesla T4 16 GB Q4_K_M 38K context · ~32 tok/s
- Tesla V100 16GB 16 GB Q4_K_M 38K context · ~90 tok/s
- RTX 4000 Ada 20 GB Q8_0 23K context · ~22 tok/s
- RTX 3090 24 GB Q8_0 45K context · ~56 tok/s
- RTX 3090 Ti 24 GB Q8_0 45K context · ~60 tok/s
- RTX 4090 24 GB Q8_0 45K context · ~60 tok/s
- RTX 4500 Ada 24 GB Q8_0 45K context · ~26 tok/s
- RTX A5000 24 GB Q8_0 45K context · ~46 tok/s
- RTX PRO 4000 Blackwell 24 GB Q8_0 45K context · ~40 tok/s
- RTX Spark 24 GB Q8_0 45K context · ~18 tok/s
- Tesla P40 24 GB Q8_0 45K context · ~21 tok/s
- RTX 5090 32 GB Q8_0 89K context · ~107 tok/s
- RTX 5000 Ada 32 GB Q8_0 89K context · ~34 tok/s
- RTX PRO 4500 Blackwell 32 GB Q8_0 89K context · ~54 tok/s
- Tesla V100 32GB 32 GB Q8_0 89K context · ~54 tok/s
- L40S 48 GB Q8_0 128K context · ~52 tok/s
- RTX 6000 Ada 48 GB Q8_0 128K context · ~57 tok/s
- RTX A6000 48 GB Q8_0 128K context · ~46 tok/s
- RTX PRO 5000 Blackwell 48 GB Q8_0 128K context · ~80 tok/s
- RTX PRO 5000 Blackwell 72GB 72 GB Q8_0 128K context · ~80 tok/s
- RTX PRO 5500 Blackwell 84 GB Q8_0 128K context · ~84 tok/s
- RTX PRO 6000 Blackwell 96 GB Q8_0 128K context · ~107 tok/s
- DGX Spark (GB 10) 128 GB Q8_0 128K context · ~16 tok/s
AMD
- Radeon RX 6700 XT 12 GB Q4_K_M 16K context · ~38 tok/s
- Radeon RX 6750 XT 12 GB Q4_K_M 16K context · ~43 tok/s
- Radeon RX 7700 XT 12 GB Q4_K_M 16K context · ~43 tok/s
- Radeon RX 9070 GRE 12 GB Q4_K_M 16K context · ~43 tok/s
- AMD BC-250 16 GB Q4_K_M 38K context · ~45 tok/s
- Radeon RX 6800 16 GB Q4_K_M 38K context · ~51 tok/s
- Radeon RX 6800 XT 16 GB Q4_K_M 38K context · ~51 tok/s
- Radeon RX 6900 XT 16 GB Q4_K_M 38K context · ~51 tok/s
- Radeon RX 6950 XT 16 GB Q4_K_M 38K context · ~58 tok/s
- Radeon RX 7600 XT 16 GB Q4_K_M 38K context · ~29 tok/s
- Radeon RX 7800 XT 16 GB Q4_K_M 38K context · ~62 tok/s
- Radeon RX 7900 GRE 16 GB Q4_K_M 38K context · ~58 tok/s
- Radeon RX 9060 XT 16GB 16 GB Q4_K_M 38K context · ~32 tok/s
- Radeon RX 9070 16 GB Q4_K_M 38K context · ~64 tok/s
- Radeon RX 9070 XT 16 GB Q4_K_M 38K context · ~64 tok/s
- Radeon RX 7900 XT 20 GB Q8_0 23K context · ~48 tok/s
- Radeon RX 7900 XTX 24 GB Q8_0 45K context · ~57 tok/s
- Radeon AI PRO R9700 32 GB Q8_0 89K context · ~38 tok/s
- Ryzen AI Max (Strix Halo) 32 GB Q8_0 89K context · ~15 tok/s
- Ryzen AI Max 400 (Gorgon Halo) 64 GB Q8_0 128K context · ~16 tok/s
Intel
- Arc B570 10 GB Q4_K_M 5K context · ~38 tok/s tight fit
- Arc B580 12 GB Q4_K_M 16K context · ~46 tok/s
- Arc A770 16GB 16 GB Q4_K_M 38K context · ~56 tok/s
- Arc Pro B65 32 GB Q8_0 89K context · ~36 tok/s
- Arc Pro B70 32 GB Q8_0 89K context · ~36 tok/s
Apple
- M1 16 GB Q4_K_M 38K context · ~7 tok/s
- M1 Pro 16 GB Q4_K_M 38K context · ~20 tok/s
- M2 16 GB Q4_K_M 38K context · ~10 tok/s
- M2 Pro 16 GB Q4_K_M 38K context · ~20 tok/s
- M3 16 GB Q4_K_M 38K context · ~10 tok/s
- M4 16 GB Q4_K_M 38K context · ~12 tok/s
- M5 16 GB Q4_K_M 38K context · ~15 tok/s
- M6 16 GB Q4_K_M 38K context · ~15 tok/s
- M3 Pro 18 GB Q4_K_M 49K context · ~15 tok/s
- M4 Pro 24 GB Q8_0 45K context · ~16 tok/s
- M5 Pro 24 GB Q8_0 45K context · ~18 tok/s
- M1 Max 32 GB Q8_0 89K context · ~24 tok/s
- M2 Max 32 GB Q8_0 89K context · ~24 tok/s
- M3 Max 36 GB Q8_0 110K context · ~18 tok/s
- M4 Max 36 GB Q8_0 110K context · ~25 tok/s
- M5 Max 36 GB Q8_0 110K context · ~28 tok/s
- M1 Ultra 64 GB Q8_0 128K context · ~48 tok/s
- M2 Ultra 64 GB Q8_0 128K context · ~48 tok/s
- M3 Ultra 96 GB Q8_0 128K context · ~49 tok/s
- M5 Ultra 96 GB Q8_0 128K context · ~72 tok/s