Hardware requirements

What hardware do I need to run Granite 4.1 3B?

3.4B 128K context Apache 2.0

At Q4_K_M the weights are 1.9 GB; with an 8K KV cache and runtime overhead that is 3.1 GB, so it wants a 8 GB card to run with headroom — in practice 8 GB VRAM: GeForce RTX 3070 Ti and 7 others. Every figure below is computed from the model’s published shape against each device’s usable memory; tokens per second are estimates. The method.

Minimum practical 8 GB VRAM

GeForce RTX 3070 Ti — Q8_0, 4.6 GB of 7.0 GB, ~109 tok/s, up to 38K context.

Also: Arc A750, GeForce RTX 5060, GeForce RTX 3070, GeForce RTX 3060 Ti, GeForce RTX 4060 Ti +2 more
the smallest card that holds it at all, at Q3_K_M — the lowest quantisation worth running
Recommended 8 GB VRAM

GeForce RTX 3070 Ti — Q4_K_M, 3.1 GB of 7.0 GB, ~193 tok/s, up to 57K context.

Also: Arc A750, GeForce RTX 5060, GeForce RTX 3070, GeForce RTX 3060 Ti, GeForce RTX 4060 Ti +2 more
runs the recommended quantisation with headroom at 8K context
High quality 8 GB VRAM

GeForce RTX 3070 Ti — Q8_0, 4.6 GB of 7.0 GB, ~109 tok/s, up to 38K context.

Also: Arc A750, GeForce RTX 5060, GeForce RTX 3070, GeForce RTX 3060 Ti, GeForce RTX 4060 Ti +2 more
near-lossless Q8_0 with headroom
Long context 20 GB VRAM

Radeon RX 7900 XT — Q4_K_M, 12.5 GB of 18.4 GB, ~253 tok/s at 128K.

Also: RTX 4000 Ada
128K tokens (or the model’s maximum) at the recommended quantisation
Apple Silicon 16 GB unified

M1 · 16 GB — Q4_K_M, 3.1 GB of 10.7 GB, ~20 tok/s, up to 104K context.

smallest Mac that runs the recommended quantisation with headroom
CPU only 32 GB RAM

CPU only · DDR4 dual-channel — Q8_0, 4.6 GB of 25.6 GB, ~6.4 tok/s, up to 128K context.

no GPU — weights stream from system RAM

How much VRAM at each quantisation

QuantWeights+KV 8K+KV 32K+KV 128KTotal @ 8KNeedsQuality
F166.3 GB0.63 GB2.50 GB10.0 GB 7.6 GB11 GB cardReference
Q8_03.4 GB0.63 GB2.50 GB10.0 GB 4.6 GB8 GB card−0.1% ppl
Q6_K2.6 GB0.63 GB2.50 GB10.0 GB 3.8 GB8 GB card−0.4% ppl
Q5_K_M2.2 GB0.63 GB2.50 GB10.0 GB 3.5 GB8 GB card−0.8% ppl
Q4_K_M1.9 GB0.63 GB2.50 GB10.0 GB 3.1 GB8 GB card−1.9% ppl
Q3_K_M1.5 GB0.63 GB2.50 GB10.0 GB 2.8 GB8 GB card−5.4% ppl
Q2_K1.3 GB0.63 GB2.50 GB10.0 GB 2.6 GB8 GB card−15% ppl

Totals add 0.6 GB runtime overhead and an f16 KV cache; "needs" is the smallest discrete card class whose usable memory (after display and driver reserve) leaves 15% headroom. A q8_0 cache roughly halves the KV columns.

Every device, checked

System RAM for offload:
163264128256

NVIDIA · GeForce

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
GeForce RTX 5090 32 GB Q8_0 Q8_0 128K 567 yes Check
GeForce RTX 5080 16 GB Q8_0 Q8_0 128K 304 tight Check
GeForce RTX 5070 Ti 16 GB Q8_0 Q8_0 128K 284 tight Check
GeForce RTX 5070 12 GB Q8_0 Q8_0 103K 213 no Check
GeForce RTX 5060 Ti 16 GB 16 GB Q8_0 Q8_0 128K 142 tight Check
GeForce RTX 5060 8 GB Q8_0 Q8_0 57K 142 no Check
GeForce RTX 4090 24 GB Q8_0 Q8_0 128K 319 yes Check
GeForce RTX 4080 Super 16 GB Q8_0 Q8_0 128K 233 tight Check
GeForce RTX 4080 16 GB Q8_0 Q8_0 128K 227 tight Check
GeForce RTX 4070 Ti Super 16 GB Q8_0 Q8_0 128K 213 tight Check
GeForce RTX 4070 Ti 12 GB Q8_0 Q8_0 103K 160 no Check
GeForce RTX 4070 Super 12 GB Q8_0 Q8_0 103K 160 no Check
GeForce RTX 4070 12 GB Q8_0 Q8_0 103K 160 no Check
GeForce RTX 4060 Ti 16 GB 16 GB Q8_0 Q8_0 128K 91 tight Check
GeForce RTX 4060 Ti 8 GB Q8_0 Q8_0 57K 91 no Check
GeForce RTX 4060 8 GB Q8_0 Q8_0 57K 86 no Check
GeForce RTX 3090 Ti 24 GB Q8_0 Q8_0 128K 319 yes Check
GeForce RTX 3090 24 GB Q8_0 Q8_0 128K 296 yes Check
GeForce RTX 3080 Ti 12 GB Q8_0 Q8_0 103K 289 no Check
GeForce RTX 3080 12 GB 12 GB Q8_0 Q8_0 103K 289 no Check
GeForce RTX 3080 10 GB Q8_0 Q8_0 80K 241 no Check
GeForce RTX 3070 Ti 8 GB Q8_0 Q8_0 57K 193 no Check
GeForce RTX 3070 8 GB Q8_0 Q8_0 57K 142 no Check
GeForce RTX 3060 Ti 8 GB Q8_0 Q8_0 57K 142 no Check
GeForce RTX 3060 12 GB 12 GB Q8_0 Q8_0 103K 114 no Check
GeForce RTX 2080 Ti 11 GB Q8_0 Q8_0 92K 195 no Check
GeForce RTX 2060 12 GB 12 GB Q8_0 Q8_0 103K 106 no Check
GeForce GTX 1080 Ti 11 GB Q8_0 Q8_0 92K 153 no Check

NVIDIA · GeForce laptop

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
GeForce RTX 5090 Laptop 24 GB Q8_0 Q8_0 128K 284 yes Check
GeForce RTX 4090 Laptop 16 GB Q8_0 Q8_0 128K 182 tight Check
GeForce RTX 4080 Laptop 12 GB Q8_0 Q8_0 103K 137 no Check
GeForce RTX 4070 Laptop 8 GB Q8_0 Q8_0 57K 81 no Check

NVIDIA · Workstation

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
RTX 6000 Ada 48 GB Q8_0 Q8_0 128K 304 yes Check
RTX 5000 Ada 32 GB Q8_0 Q8_0 128K 182 yes Check
RTX 4000 Ada 20 GB Q8_0 Q8_0 128K 114 yes Check
RTX 2000 Ada 16 GB Q8_0 Q8_0 128K 71 tight Check
RTX A6000 48 GB Q8_0 Q8_0 128K 243 yes Check
RTX A5000 24 GB Q8_0 Q8_0 128K 243 yes Check
RTX A4000 16 GB Q8_0 Q8_0 128K 142 tight Check

NVIDIA · Data centre

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
L40S 48 GB Q8_0 Q8_0 128K 274 yes Check
L4 24 GB Q8_0 Q8_0 128K 95 yes Check
A10 24 GB Q8_0 Q8_0 128K 190 yes Check
A100 40 GB 40 GB Q8_0 Q8_0 128K 492 yes Check
A100 80 GB 80 GB Q8_0 Q8_0 128K 646 yes Check
H100 PCIe 80 GB Q8_0 Q8_0 128K 633 yes Check
H100 SXM 80 GB Q8_0 Q8_0 128K 1061 yes Check
H200 SXM 141 GB Q8_0 Q8_0 128K 1520 yes Check
Tesla P40 24 GB Q8_0 Q8_0 128K 110 yes Check

AMD · Radeon

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
Radeon RX 9070 XT 16 GB Q8_0 Q8_0 128K 204 tight Check
Radeon RX 9070 16 GB Q8_0 Q8_0 128K 204 tight Check
Radeon RX 7900 XTX 24 GB Q8_0 Q8_0 128K 304 yes Check
Radeon RX 7900 XT 20 GB Q8_0 Q8_0 128K 253 yes Check
Radeon RX 7900 GRE 16 GB Q8_0 Q8_0 128K 182 tight Check
Radeon RX 7800 XT 16 GB Q8_0 Q8_0 128K 198 tight Check
Radeon RX 7700 XT 12 GB Q8_0 Q8_0 103K 137 no Check
Radeon RX 7600 XT 16 GB Q8_0 Q8_0 128K 91 tight Check
Radeon RX 6900 XT 16 GB Q8_0 Q8_0 128K 162 tight Check
Radeon RX 6800 XT 16 GB Q8_0 Q8_0 128K 162 tight Check
Radeon RX 6700 XT 12 GB Q8_0 Q8_0 103K 122 no Check

AMD · Radeon Pro

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
Radeon Pro W7900 48 GB Q8_0 Q8_0 128K 274 yes Check
Radeon Pro W7800 32 GB Q8_0 Q8_0 128K 182 yes Check

AMD · Instinct

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
Instinct MI210 64 GB Q8_0 Q8_0 128K 519 yes Check
Instinct MI300X 192 GB Q8_0 Q8_0 128K 1678 yes Check

AMD · Ryzen AI Max

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
Ryzen AI Max+ 395 · 128 GB 128 GB Q8_0 Q8_0 128K 75 yes Check
Ryzen AI Max+ 395 · 64 GB 64 GB Q8_0 Q8_0 128K 75 yes Check

Apple · Apple Silicon

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
M1 · 16 GB 16 GB Q8_0 Q8_0 104K 20 no Check
M1 Pro · 32 GB 32 GB Q8_0 Q8_0 128K 58 yes Check
M1 Max · 64 GB 64 GB Q8_0 Q8_0 128K 117 yes Check
M1 Ultra · 128 GB 128 GB Q8_0 Q8_0 128K 234 yes Check
M2 · 24 GB 24 GB Q8_0 Q8_0 128K 29 yes Check
M2 Pro · 32 GB 32 GB Q8_0 Q8_0 128K 58 yes Check
M2 Max · 96 GB 96 GB Q8_0 Q8_0 128K 117 yes Check
M2 Ultra · 192 GB 192 GB Q8_0 Q8_0 128K 234 yes Check
M3 · 24 GB 24 GB Q8_0 Q8_0 128K 29 yes Check
M3 Pro · 36 GB 36 GB Q8_0 Q8_0 128K 44 yes Check
M3 Max · 64 GB 64 GB Q8_0 Q8_0 128K 117 yes Check
M3 Max · 128 GB 128 GB Q8_0 Q8_0 128K 117 yes Check
M3 Ultra · 256 GB 256 GB Q8_0 Q8_0 128K 234 yes Check
M3 Ultra · 512 GB 512 GB Q8_0 Q8_0 128K 234 yes Check
M4 · 32 GB 32 GB Q8_0 Q8_0 128K 35 yes Check
M4 Pro · 48 GB 48 GB Q8_0 Q8_0 128K 80 yes Check
M4 Pro · 64 GB 64 GB Q8_0 Q8_0 128K 80 yes Check
M4 Max · 64 GB 64 GB Q8_0 Q8_0 128K 120 yes Check
M4 Max · 128 GB 128 GB Q8_0 Q8_0 128K 160 yes Check

Intel · Arc

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
Arc B580 12 GB Q8_0 Q8_0 103K 144 no Check
Arc B570 10 GB Q8_0 Q8_0 80K 120 no Check
Arc A770 16 GB 16 GB Q8_0 Q8_0 128K 177 tight Check
Arc A750 8 GB Q8_0 Q8_0 57K 162 no Check

CPU · CPU + system RAM

DeviceMemoryBest quant with headroomFits at allMax context @ Q4_K_MTok/s est.128K
CPU only · DDR4 dual-channel 32 GB Q8_0 Q8_0 128K 11 yes Check
CPU only · DDR5 dual-channel 64 GB Q8_0 Q8_0 128K 20 yes Check
CPU only · DDR5 8-channel server 256 GB Q8_0 Q8_0 128K 67 yes Check