What GPU do you need to run Mistral Nemo 12B?

12.2B params 128K max context Tool use

The smallest GPU memory that runs Mistral Nemo 12B well is 11 GB (Q4_K_M, e.g. GeForce RTX 2080 Ti). 88 of 132 GPUs run it well and 32 more are a tight fit.

MMLU 68

VRAM needed per quantization

Quantization	Weights	16K context	32K context	Full 128K
Q4_K_M	7.6 GB	10.7 GB	13.7 GB	32.0 GB
Q8_0	13.1 GB	16.2 GB	19.2 GB	37.5 GB

Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.

Cheapest GPUs that can run Mistral Nemo 12B

Lowest launch MSRP among GPUs that run its smallest quantization (Q4_K_M) well, with at least 16K context.

Arc B580 12 GB · ~53 tok/s at Q4_K_M · Launch MSRP $249
Check price on Amazon
Radeon RX 7600 XT 16 GB · ~33 tok/s at Q4_K_M · Launch MSRP $329
Check price on Amazon
RTX 3060 12 GB · ~42 tok/s at Q4_K_M · Launch MSRP $329
Check price on Amazon

As an Amazon Associate, I earn from qualifying purchases.

GPUs that can run Mistral Nemo 12B

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 1070 8 GB Q4_K_M 2K context · ~30 tok/s tight fit
GTX 1070 Ti 8 GB Q4_K_M 2K context · ~30 tok/s tight fit
GTX 1080 8 GB Q4_K_M 2K context · ~37 tok/s tight fit
RTX 2060 SUPER 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
RTX 2070 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
RTX 2070 SUPER 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
RTX 2080 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
RTX 2080 SUPER 8 GB Q4_K_M 2K context · ~58 tok/s tight fit
RTX 3050 8 GB Q4_K_M 2K context · ~26 tok/s tight fit
RTX 3060 Ti 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
RTX 3060 Ti GDDR6X 8 GB Q4_K_M 2K context · ~71 tok/s tight fit
RTX 3070 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
RTX 3070 Ti 8 GB Q4_K_M 2K context · ~71 tok/s tight fit
RTX 4060 8 GB Q4_K_M 2K context · ~32 tok/s tight fit
RTX 4060 Ti 8 GB Q4_K_M 2K context · ~33 tok/s tight fit
RTX 5050 8 GB Q4_K_M 2K context · ~37 tok/s tight fit
RTX 5060 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
RTX 5060 Ti 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
T1000 8GB 8 GB Q4_K_M 2K context · ~19 tok/s tight fit
RTX 3080 10 GB Q4_K_M 12K context · ~88 tok/s tight fit
GTX 1080 Ti 11 GB Q4_K_M 17K context · ~56 tok/s
RTX 2080 Ti 11 GB Q4_K_M 17K context · ~72 tok/s
RTX 2060 12GB 12 GB Q4_K_M 23K context · ~39 tok/s
RTX 3060 12 GB Q4_K_M 23K context · ~42 tok/s
RTX 3080 12GB 12 GB Q4_K_M 23K context · ~106 tok/s
RTX 3080 Ti 12 GB Q4_K_M 23K context · ~106 tok/s
RTX 4070 12 GB Q4_K_M 23K context · ~59 tok/s
RTX 4070 SUPER 12 GB Q4_K_M 23K context · ~59 tok/s
RTX 4070 Ti 12 GB Q4_K_M 23K context · ~59 tok/s
RTX 5070 12 GB Q4_K_M 23K context · ~78 tok/s
RTX A2000 12GB 12 GB Q4_K_M 23K context · ~33 tok/s
RTX 4060 Ti 16GB 16 GB Q4_K_M 43K context · ~33 tok/s
RTX 4070 Ti SUPER 16 GB Q4_K_M 43K context · ~78 tok/s
RTX 4080 16 GB Q4_K_M 43K context · ~83 tok/s
RTX 4080 SUPER 16 GB Q4_K_M 43K context · ~86 tok/s
RTX 5060 Ti 16GB 16 GB Q4_K_M 43K context · ~52 tok/s
RTX 5070 Ti 16 GB Q4_K_M 43K context · ~104 tok/s
RTX 5080 16 GB Q4_K_M 43K context · ~112 tok/s
RTX 2000 Ada 16 GB Q4_K_M 43K context · ~26 tok/s
RTX A4000 16 GB Q4_K_M 43K context · ~52 tok/s
RTX PRO 2000 Blackwell 16 GB Q4_K_M 43K context · ~33 tok/s
Tesla T4 16 GB Q4_K_M 43K context · ~37 tok/s
Tesla V100 16GB 16 GB Q4_K_M 43K context · ~105 tok/s
RTX 4000 Ada 20 GB Q8_0 36K context · ~26 tok/s
RTX 3090 24 GB Q8_0 57K context · ~66 tok/s
RTX 3090 Ti 24 GB Q8_0 57K context · ~71 tok/s
RTX 4090 24 GB Q8_0 57K context · ~71 tok/s
RTX 4500 Ada 24 GB Q8_0 57K context · ~31 tok/s
RTX A5000 24 GB Q8_0 57K context · ~54 tok/s
RTX PRO 4000 Blackwell 24 GB Q8_0 57K context · ~48 tok/s
RTX Spark 24 GB Q8_0 57K context · ~21 tok/s
Tesla P40 24 GB Q8_0 57K context · ~25 tok/s
RTX 5090 32 GB Q8_0 98K context · ~127 tok/s
RTX 5000 Ada 32 GB Q8_0 98K context · ~41 tok/s
RTX PRO 4500 Blackwell 32 GB Q8_0 98K context · ~64 tok/s
Tesla V100 32GB 32 GB Q8_0 98K context · ~64 tok/s
L40S 48 GB Q8_0 128K context · ~61 tok/s
RTX 6000 Ada 48 GB Q8_0 128K context · ~68 tok/s
RTX A6000 48 GB Q8_0 128K context · ~54 tok/s
RTX PRO 5000 Blackwell 48 GB Q8_0 128K context · ~95 tok/s
RTX PRO 5000 Blackwell 72GB 72 GB Q8_0 128K context · ~95 tok/s
RTX PRO 5500 Blackwell 84 GB Q8_0 128K context · ~99 tok/s
RTX PRO 6000 Blackwell 96 GB Q8_0 128K context · ~127 tok/s
DGX Spark (GB 10) 128 GB Q8_0 128K context · ~19 tok/s

AMD

Radeon RX 5700 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
Radeon RX 5700 XT 8 GB Q4_K_M 2K context · ~52 tok/s tight fit
Radeon RX 6600 8 GB Q4_K_M 2K context · ~26 tok/s tight fit
Radeon RX 6600 XT 8 GB Q4_K_M 2K context · ~30 tok/s tight fit
Radeon RX 6650 XT 8 GB Q4_K_M 2K context · ~33 tok/s tight fit
Radeon RX 7600 8 GB Q4_K_M 2K context · ~33 tok/s tight fit
Radeon RX 9060 8 GB Q4_K_M 2K context · ~33 tok/s tight fit
Radeon RX 9060 XT 8 GB Q4_K_M 2K context · ~37 tok/s tight fit
Radeon RX 6700 XT 12 GB Q4_K_M 23K context · ~45 tok/s
Radeon RX 6750 XT 12 GB Q4_K_M 23K context · ~50 tok/s
Radeon RX 7700 XT 12 GB Q4_K_M 23K context · ~50 tok/s
Radeon RX 9070 GRE 12 GB Q4_K_M 23K context · ~50 tok/s
AMD BC-250 16 GB Q4_K_M 43K context · ~52 tok/s
Radeon RX 6800 16 GB Q4_K_M 43K context · ~60 tok/s
Radeon RX 6800 XT 16 GB Q4_K_M 43K context · ~60 tok/s
Radeon RX 6900 XT 16 GB Q4_K_M 43K context · ~60 tok/s
Radeon RX 6950 XT 16 GB Q4_K_M 43K context · ~67 tok/s
Radeon RX 7600 XT 16 GB Q4_K_M 43K context · ~33 tok/s
Radeon RX 7800 XT 16 GB Q4_K_M 43K context · ~73 tok/s
Radeon RX 7900 GRE 16 GB Q4_K_M 43K context · ~67 tok/s
Radeon RX 9060 XT 16GB 16 GB Q4_K_M 43K context · ~37 tok/s
Radeon RX 9070 16 GB Q4_K_M 43K context · ~74 tok/s
Radeon RX 9070 XT 16 GB Q4_K_M 43K context · ~74 tok/s
Radeon RX 7900 XT 20 GB Q8_0 36K context · ~57 tok/s
Radeon RX 7900 XTX 24 GB Q8_0 57K context · ~68 tok/s
Radeon AI PRO R9700 32 GB Q8_0 98K context · ~45 tok/s
Ryzen AI Max (Strix Halo) 32 GB Q8_0 98K context · ~18 tok/s
Ryzen AI Max 400 (Gorgon Halo) 64 GB Q8_0 128K context · ~19 tok/s

Intel

Arc A580 8 GB Q4_K_M 2K context · ~60 tok/s tight fit
Arc A750 8 GB Q4_K_M 2K context · ~60 tok/s tight fit
Arc A770 8GB 8 GB Q4_K_M 2K context · ~60 tok/s tight fit
Arc B570 10 GB Q4_K_M 12K context · ~44 tok/s tight fit
Arc B580 12 GB Q4_K_M 23K context · ~53 tok/s
Arc A770 16GB 16 GB Q4_K_M 43K context · ~65 tok/s
Arc Pro B65 32 GB Q8_0 98K context · ~43 tok/s
Arc Pro B70 32 GB Q8_0 98K context · ~43 tok/s

Apple

M1 16 GB Q4_K_M 43K context · ~8 tok/s
M1 Pro 16 GB Q4_K_M 43K context · ~23 tok/s
M2 16 GB Q4_K_M 43K context · ~12 tok/s
M2 Pro 16 GB Q4_K_M 43K context · ~23 tok/s
M3 16 GB Q4_K_M 43K context · ~12 tok/s
M4 16 GB Q4_K_M 43K context · ~14 tok/s
M5 16 GB Q4_K_M 43K context · ~18 tok/s
M6 16 GB Q4_K_M 43K context · ~18 tok/s
M3 Pro 18 GB Q8_0 25K context · ~11 tok/s
M4 Pro 24 GB Q8_0 57K context · ~19 tok/s
M5 Pro 24 GB Q8_0 57K context · ~22 tok/s
M1 Max 32 GB Q8_0 98K context · ~28 tok/s
M2 Max 32 GB Q8_0 98K context · ~28 tok/s
M3 Max 36 GB Q8_0 119K context · ~21 tok/s
M4 Max 36 GB Q8_0 119K context · ~29 tok/s
M5 Max 36 GB Q8_0 119K context · ~33 tok/s
M1 Ultra 64 GB Q8_0 128K context · ~57 tok/s
M2 Ultra 64 GB Q8_0 128K context · ~57 tok/s
M3 Ultra 96 GB Q8_0 128K context · ~58 tok/s
M5 Ultra 96 GB Q8_0 128K context · ~85 tok/s

Check your exact setup in the full tool →