What GPU do you need to run DeepSeek Coder 6.7B?
6.74B params 16K max context
The smallest GPU memory that runs DeepSeek Coder 6.7B well is 12 GB (Q4_K_M, e.g. GeForce RTX 5070). 86 of 132 GPUs run it well and 42 more are a tight fit.
MMLU 50
VRAM needed per quantization
| Quantization | Weights | 16K context | Full 16K |
|---|---|---|---|
| Q4_K_M | 4.1 GB | 11.9 GB | 11.9 GB |
| Q8_0 | 7.2 GB | 15.0 GB | 15.0 GB |
Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.
Cheapest GPUs that can run DeepSeek Coder 6.7B
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 DeepSeek Coder 6.7B
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
- GTX 1060 6GB 6 GB Q4_K_M 3K context · ~38 tok/s tight fit
- GTX 1660 6 GB Q4_K_M 3K context · ~38 tok/s tight fit
- GTX 1660 SUPER 6 GB Q4_K_M 3K context · ~66 tok/s tight fit
- GTX 1660 Ti 6 GB Q4_K_M 3K context · ~57 tok/s tight fit
- RTX 2060 6 GB Q4_K_M 3K context · ~66 tok/s tight fit
- RTX 3050 6GB 6 GB Q4_K_M 3K context · ~33 tok/s tight fit
- RTX A2000 6 GB Q4_K_M 3K context · ~57 tok/s tight fit
- GTX 1070 8 GB Q8_0 1K context · ~31 tok/s tight fit
- GTX 1070 Ti 8 GB Q8_0 1K context · ~31 tok/s tight fit
- GTX 1080 8 GB Q8_0 1K context · ~39 tok/s tight fit
- RTX 2060 SUPER 8 GB Q8_0 1K context · ~55 tok/s tight fit
- RTX 2070 8 GB Q8_0 1K context · ~55 tok/s tight fit
- RTX 2070 SUPER 8 GB Q8_0 1K context · ~55 tok/s tight fit
- RTX 2080 8 GB Q8_0 1K context · ~55 tok/s tight fit
- RTX 2080 SUPER 8 GB Q8_0 1K context · ~61 tok/s tight fit
- RTX 3050 8 GB Q8_0 1K context · ~27 tok/s tight fit
- RTX 3060 Ti 8 GB Q8_0 1K context · ~55 tok/s tight fit
- RTX 3060 Ti GDDR6X 8 GB Q8_0 1K context · ~75 tok/s tight fit
- RTX 3070 8 GB Q8_0 1K context · ~55 tok/s tight fit
- RTX 3070 Ti 8 GB Q8_0 1K context · ~75 tok/s tight fit
- RTX 4060 8 GB Q8_0 1K context · ~33 tok/s tight fit
- RTX 4060 Ti 8 GB Q8_0 1K context · ~35 tok/s tight fit
- RTX 5050 8 GB Q8_0 1K context · ~39 tok/s tight fit
- RTX 5060 8 GB Q8_0 1K context · ~55 tok/s tight fit
- RTX 5060 Ti 8 GB Q8_0 1K context · ~55 tok/s tight fit
- T1000 8GB 8 GB Q8_0 1K context · ~20 tok/s tight fit
- RTX 3080 10 GB Q8_0 5K context · ~93 tok/s tight fit
- GTX 1080 Ti 11 GB Q8_0 7K context · ~59 tok/s tight fit
- RTX 2080 Ti 11 GB Q8_0 7K context · ~75 tok/s tight fit
- RTX 2060 12GB 12 GB Q4_K_M 16K context · ~66 tok/s
- RTX 3060 12 GB Q4_K_M 16K context · ~71 tok/s
- RTX 3080 12GB 12 GB Q4_K_M 16K context · ~180 tok/s
- RTX 3080 Ti 12 GB Q4_K_M 16K context · ~180 tok/s
- RTX 4070 12 GB Q4_K_M 16K context · ~99 tok/s
- RTX 4070 SUPER 12 GB Q4_K_M 16K context · ~99 tok/s
- RTX 4070 Ti 12 GB Q4_K_M 16K context · ~99 tok/s
- RTX 5070 12 GB Q4_K_M 16K context · ~132 tok/s
- RTX A2000 12GB 12 GB Q4_K_M 16K context · ~57 tok/s
- RTX 4060 Ti 16GB 16 GB Q8_0 16K context · ~35 tok/s
- RTX 4070 Ti SUPER 16 GB Q8_0 16K context · ~82 tok/s
- RTX 4080 16 GB Q8_0 16K context · ~88 tok/s
- RTX 4080 SUPER 16 GB Q8_0 16K context · ~90 tok/s
- RTX 5060 Ti 16GB 16 GB Q8_0 16K context · ~55 tok/s
- RTX 5070 Ti 16 GB Q8_0 16K context · ~110 tok/s
- RTX 5080 16 GB Q8_0 16K context · ~118 tok/s
- RTX 2000 Ada 16 GB Q8_0 16K context · ~27 tok/s
- RTX A4000 16 GB Q8_0 16K context · ~55 tok/s
- RTX PRO 2000 Blackwell 16 GB Q8_0 16K context · ~35 tok/s
- Tesla T4 16 GB Q8_0 16K context · ~39 tok/s
- Tesla V100 16GB 16 GB Q8_0 16K context · ~110 tok/s
- RTX 4000 Ada 20 GB Q8_0 16K context · ~44 tok/s
- RTX 3090 24 GB Q8_0 16K context · ~115 tok/s
- RTX 3090 Ti 24 GB Q8_0 16K context · ~124 tok/s
- RTX 4090 24 GB Q8_0 16K context · ~124 tok/s
- RTX 4500 Ada 24 GB Q8_0 16K context · ~53 tok/s
- RTX A5000 24 GB Q8_0 16K context · ~94 tok/s
- RTX PRO 4000 Blackwell 24 GB Q8_0 16K context · ~82 tok/s
- RTX Spark 24 GB Q8_0 16K context · ~37 tok/s
- Tesla P40 24 GB Q8_0 16K context · ~42 tok/s
- RTX 5090 32 GB Q8_0 16K context · ~220 tok/s
- RTX 5000 Ada 32 GB Q8_0 16K context · ~71 tok/s
- RTX PRO 4500 Blackwell 32 GB Q8_0 16K context · ~110 tok/s
- Tesla V100 32GB 32 GB Q8_0 16K context · ~110 tok/s
- L40S 48 GB Q8_0 16K context · ~106 tok/s
- RTX 6000 Ada 48 GB Q8_0 16K context · ~118 tok/s
- RTX A6000 48 GB Q8_0 16K context · ~94 tok/s
- RTX PRO 5000 Blackwell 48 GB Q8_0 16K context · ~165 tok/s
- RTX PRO 5000 Blackwell 72GB 72 GB Q8_0 16K context · ~165 tok/s
- RTX PRO 5500 Blackwell 84 GB Q8_0 16K context · ~171 tok/s
- RTX PRO 6000 Blackwell 96 GB Q8_0 16K context · ~220 tok/s
- DGX Spark (GB 10) 128 GB Q8_0 16K context · ~33 tok/s
AMD
- Radeon RX 5600 XT 6 GB Q4_K_M 3K context · ~57 tok/s tight fit
- Radeon RX 5700 8 GB Q8_0 1K context · ~55 tok/s tight fit
- Radeon RX 5700 XT 8 GB Q8_0 1K context · ~55 tok/s tight fit
- Radeon RX 6600 8 GB Q8_0 1K context · ~27 tok/s tight fit
- Radeon RX 6600 XT 8 GB Q8_0 1K context · ~31 tok/s tight fit
- Radeon RX 6650 XT 8 GB Q8_0 1K context · ~34 tok/s tight fit
- Radeon RX 7600 8 GB Q8_0 1K context · ~35 tok/s tight fit
- Radeon RX 9060 8 GB Q8_0 1K context · ~35 tok/s tight fit
- Radeon RX 9060 XT 8 GB Q8_0 1K context · ~39 tok/s tight fit
- Radeon RX 6700 XT 12 GB Q4_K_M 16K context · ~76 tok/s
- Radeon RX 6750 XT 12 GB Q4_K_M 16K context · ~85 tok/s
- Radeon RX 7700 XT 12 GB Q4_K_M 16K context · ~85 tok/s
- Radeon RX 9070 GRE 12 GB Q4_K_M 16K context · ~85 tok/s
- AMD BC-250 16 GB Q8_0 16K context · ~55 tok/s
- Radeon RX 6800 16 GB Q8_0 16K context · ~63 tok/s
- Radeon RX 6800 XT 16 GB Q8_0 16K context · ~63 tok/s
- Radeon RX 6900 XT 16 GB Q8_0 16K context · ~63 tok/s
- Radeon RX 6950 XT 16 GB Q8_0 16K context · ~71 tok/s
- Radeon RX 7600 XT 16 GB Q8_0 16K context · ~35 tok/s
- Radeon RX 7800 XT 16 GB Q8_0 16K context · ~76 tok/s
- Radeon RX 7900 GRE 16 GB Q8_0 16K context · ~71 tok/s
- Radeon RX 9060 XT 16GB 16 GB Q8_0 16K context · ~39 tok/s
- Radeon RX 9070 16 GB Q8_0 16K context · ~78 tok/s
- Radeon RX 9070 XT 16 GB Q8_0 16K context · ~78 tok/s
- Radeon RX 7900 XT 20 GB Q8_0 16K context · ~98 tok/s
- Radeon RX 7900 XTX 24 GB Q8_0 16K context · ~118 tok/s
- Radeon AI PRO R9700 32 GB Q8_0 16K context · ~78 tok/s
- Ryzen AI Max (Strix Halo) 32 GB Q8_0 16K context · ~31 tok/s
- Ryzen AI Max 400 (Gorgon Halo) 64 GB Q8_0 16K context · ~33 tok/s
Intel
- Arc A580 8 GB Q8_0 1K context · ~63 tok/s tight fit
- Arc A750 8 GB Q8_0 1K context · ~63 tok/s tight fit
- Arc A770 8GB 8 GB Q8_0 1K context · ~63 tok/s tight fit
- Arc B570 10 GB Q8_0 5K context · ~47 tok/s tight fit
- Arc B580 12 GB Q4_K_M 16K context · ~90 tok/s
- Arc A770 16GB 16 GB Q8_0 16K context · ~69 tok/s
- Arc Pro B65 32 GB Q8_0 16K context · ~75 tok/s
- Arc Pro B70 32 GB Q8_0 16K context · ~75 tok/s
Apple
- M1 16 GB Q8_0 16K context · ~8 tok/s
- M1 Pro 16 GB Q8_0 16K context · ~25 tok/s
- M2 16 GB Q8_0 16K context · ~12 tok/s
- M2 Pro 16 GB Q8_0 16K context · ~25 tok/s
- M3 16 GB Q8_0 16K context · ~12 tok/s
- M4 16 GB Q8_0 16K context · ~15 tok/s
- M5 16 GB Q8_0 16K context · ~19 tok/s
- M6 16 GB Q8_0 16K context · ~19 tok/s
- M3 Pro 18 GB Q8_0 16K context · ~18 tok/s
- M4 Pro 24 GB Q8_0 16K context · ~33 tok/s
- M5 Pro 24 GB Q8_0 16K context · ~38 tok/s
- M1 Max 32 GB Q8_0 16K context · ~49 tok/s
- M2 Max 32 GB Q8_0 16K context · ~49 tok/s
- M3 Max 36 GB Q8_0 16K context · ~37 tok/s
- M4 Max 36 GB Q8_0 16K context · ~50 tok/s
- M5 Max 36 GB Q8_0 16K context · ~56 tok/s
- M1 Ultra 64 GB Q8_0 16K context · ~98 tok/s
- M2 Ultra 64 GB Q8_0 16K context · ~98 tok/s
- M3 Ultra 96 GB Q8_0 16K context · ~100 tok/s
- M5 Ultra 96 GB Q8_0 16K context · ~147 tok/s