171 devices · 11 brands

Local AI hardware

Every GPU, laptop, mini-PC and AI station we track for local LLMs, ranked by the memory that decides what fits and the bandwidth that decides how fast it runs.

Specs verified 2026-10-07 · Prices verified 2026-09-29

Devices17111 brands

Most usable memory384 GBApple M5 Ultra

Fastest memory3,350 GB/sNVIDIA H100 80GB (SXM)

Lowest $ per GB (16 GB+)$21.81AMD Radeon RX 9060 XT 16GB

Browse by type

95 devicesGPU databaseMost memory: Apple M3 Ultra 43 devicesLaptopsMost memory: ASUS ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB)33 devicesAI stationsMost memory: Mac Studio (M3 Ultra, 256 GB)

Device	Usable memory	Bandwidth	8B at Q4	Fit	Largest model at Q4	Price
Apple M3 UltraApple Silicon · ARM, 3nm TSMC · 2025	384 GB unified	819 GB/s	43–90 tok/s	169	Nemotron 3 Ultra 550B · 339.2 GB	—
Apple M5 UltraApple Silicon · ARM, 3nm TSMC · 2026	384 GB unified	1,229 GB/s	59–123 tok/s	169	Nemotron 3 Ultra 550B · 339.2 GB	—
Mac Studio (M3 Ultra, 256 GB)AI station · ARM, 3nm TSMC · 2025	192 GB unified	819 GB/s	43–90 tok/s	159	MiMo-V2.6-Flash · 188.4 GB	—
Mac Studio (M5 Ultra, 256 GB)AI station · ARM, 3nm TSMC · 2026	192 GB unified	1,229 GB/s	59–123 tok/s	159	MiMo-V2.6-Flash · 188.4 GB	—
Apple M2 UltraApple Silicon · ARM, 5nm TSMC · 2023	144 GB unified	800 GB/s	43–89 tok/s	156	Qwen 3 235B-A22B (MoE) · 142.7 GB	—
Mac Studio (M2 Ultra, 192 GB)AI station · ARM, 5nm TSMC · 2023	144 GB unified	800 GB/s	43–89 tok/s	156	Qwen 3 235B-A22B (MoE) · 142.7 GB	—
AMD Ryzen AI Max+ 395SoC / APU · Zen 5 + RDNA 3.5 "Strix Halo" · 2025	128 GB unified	256 GB/s	18–38 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
ASUS Ascent GX10 (GB10, 128 GB)Mini PC · GB10 Grace Blackwell Superchip · 2025	128 GB unified	273 GB/s	26–53 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
ASUS ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB)Laptop · Zen 5 + RDNA 3.5 "Strix Halo" · 2025	128 GB unified	256 GB/s	18–38 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB)Mini PC · Zen 5 + RDNA 3.5 "Strix Halo" · 2025	128 GB unified	256 GB/s	18–38 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
Framework Desktop (Ryzen AI Max+ 395, 128 GB)Mini PC · Zen 5 + RDNA 3.5 "Strix Halo" · 2025	128 GB unified	256 GB/s	18–38 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB)Mini PC · Zen 5 + RDNA 3.5 "Strix Halo" · 2025	128 GB unified	256 GB/s	18–38 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
HP Z2 Mini G1a (Ryzen AI Max+ PRO 395, 128 GB)Mini PC · Zen 5 + RDNA 3.5 "Strix Halo" · 2025	128 GB unified	256 GB/s	18–38 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
HP ZBook Ultra G1a (Ryzen AI Max+ PRO 395, 128 GB)Laptop · Zen 5 + RDNA 3.5 "Strix Halo" · 2025	128 GB unified	256 GB/s	18–38 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
HP ZGX Nano G1n AI Station (GB10, 128 GB)Mini PC · GB10 Grace Blackwell Superchip · 2025	128 GB unified	273 GB/s	26–53 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
NVIDIA DGX SparkAI station · GB10 Grace Blackwell Superchip · 2025	128 GB unified	273 GB/s	26–53 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	$4,699
NVIDIA DGX Spark (128 GB)Mini PC · GB10 Grace Blackwell Superchip · 2025	128 GB unified	273 GB/s	26–53 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
NVIDIA GB10 Grace BlackwellSoC / APU · GB10 Grace Blackwell Superchip · 2025	128 GB unified	273 GB/s	26–53 tok/s	153	Qwen3.8-Flash-Next · 109.5 GB	—
Apple M1 UltraApple Silicon · ARM, 5nm TSMC · 2022	96 GB unified	800 GB/s	43–89 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
Apple M3 MaxApple Silicon · ARM, 3nm TSMC · 2023	96 GB unified	400 GB/s	24–49 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
Apple M4 MaxApple Silicon · ARM, 3nm TSMC · 2024	96 GB unified	546 GB/s	31–64 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
Apple M5 MaxApple Silicon · ARM, 3nm TSMC · 2026	96 GB unified	614 GB/s	34–71 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
Mac Studio (M5 Max 40-core GPU, 128 GB)AI station · ARM, 3nm TSMC · 2026	96 GB unified	614 GB/s	34–71 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
MacBook Pro 16" (M4 Max, 128 GB)Laptop · ARM, 3nm TSMC · 2024	96 GB unified	546 GB/s	31–64 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
MacBook Pro 16" (M5 Max, 128 GB)Laptop · ARM, 3nm TSMC · 2026	96 GB unified	614 GB/s	34–71 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
NVIDIA RTX PRO 6000 BlackwellWorkstation GPU · Blackwell GB202 · 2025	96 GB	1,792 GB/s	113–234 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	$8,565
NVIDIA RTX PRO 6000 Blackwell (Max-Q Edition)Workstation GPU · Blackwell GB202 · 2025-03-18	96 GB	1,792 GB/s	113–234 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
NVIDIA A100 80GB (PCIe)Datacenter GPU · Ampere GA100 · 2021	80 GB	1,935 GB/s	102–196 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	$15,000
NVIDIA H100 80GB (PCIe)Datacenter GPU · Hopper GH100 · 2022	80 GB	2,000 GB/s	120–250 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	$25,000
NVIDIA H100 80GB (SXM)Datacenter GPU · Hopper GH100 · 2022-09-20	80 GB	3,350 GB/s	157–327 tok/s	152	Mistral Medium 3.5 128B · 77.9 GB	—
Apple M2 MaxApple Silicon · ARM, 5nm TSMC · 2023	72 GB unified	400 GB/s	24–49 tok/s	147	GPT-OSS 120B · 71.3 GB	—
Apple M3 Max (30-core GPU)Apple Silicon · ARM, 3nm TSMC · 2023	72 GB unified	300 GB/s	18–38 tok/s	147	GPT-OSS 120B · 71.3 GB	—
Mac Studio (M5 Ultra, 96 GB)AI station · ARM, 3nm TSMC · 2026	72 GB unified	1,229 GB/s	59–123 tok/s	147	GPT-OSS 120B · 71.3 GB	—
NVIDIA RTX PRO 5000 Blackwell (72 GB)Workstation GPU · Blackwell GB202 · 2025-03-18	72 GB	1,344 GB/s	94–194 tok/s	147	GPT-OSS 120B · 71.3 GB	—
Apple M1 MaxApple Silicon · ARM, 5nm TSMC · 2021	48 GB unified	400 GB/s	24–49 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
Apple M5 ProApple Silicon · ARM, 3nm TSMC · 2026	48 GB unified	307 GB/s	19–39 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
Mac mini (M4 Pro, 64 GB)Mini PC · ARM, 3nm TSMC · 2024	48 GB unified	273 GB/s	17–35 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
Mac mini (M5 Pro, 64 GB)Mini PC · ARM, 3nm TSMC · 2026	48 GB unified	307 GB/s	19–39 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
Mac Studio (M4 Max, 64 GB)AI station · ARM, 3nm TSMC · 2025	48 GB unified	546 GB/s	31–64 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
MacBook Pro 14" (M5 Pro, 64 GB)Laptop · ARM, 3nm TSMC · 2026	48 GB unified	307 GB/s	19–39 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
MacBook Pro 16" (M5 Max 40-core GPU, 64 GB)Laptop · ARM, 3nm TSMC · 2026	48 GB unified	614 GB/s	34–71 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
MacBook Pro 16" (M5 Pro, 64 GB)Laptop · ARM, 3nm TSMC · 2026	48 GB unified	307 GB/s	19–39 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
NVIDIA L40Datacenter GPU · Ada Lovelace AD102 · 2022-10-13	48 GB	864 GB/s	77–147 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
NVIDIA L40SDatacenter GPU · Ada Lovelace AD102 · 2023	48 GB	864 GB/s	77–147 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	$7,499
AMD Radeon PRO W7900Workstation GPU · RDNA 3 Navi 31 · 2023-04-13	48 GB	864 GB/s	50–103 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
NVIDIA RTX 6000 Ada GenerationWorkstation GPU · Ada Lovelace AD102 · 2022	48 GB	960 GB/s	83–159 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	$6,799
NVIDIA RTX A6000Workstation GPU · Ampere GA102 · 2020-10-05used	48 GB	768 GB/s	51–98 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	$4,649
NVIDIA RTX PRO 5000 Blackwell (48 GB)Workstation GPU · Blackwell GB202 · 2025-03-18	48 GB	1,344 GB/s	94–194 tok/s	139	Qwen 2.5 VL 72B Instruct · 45.1 GB	—
Mac mini (M5 Pro, 48 GB)Mini PC · ARM, 3nm TSMC · 2026	36 GB unified	307 GB/s	19–39 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
MacBook Pro 14" (M5 Pro, 48 GB)Laptop · ARM, 3nm TSMC · 2026	36 GB unified	307 GB/s	19–39 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
MacBook Pro 16" (M5 Pro, 48 GB)Laptop · ARM, 3nm TSMC · 2026	36 GB unified	307 GB/s	19–39 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
MacBook Pro 16" (M4 Max, 48 GB)Laptop · ARM, 3nm TSMC · 2024	36 GB unified	546 GB/s	31–64 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$3,999
AMD Instinct MI50 32GBDatacenter GPU · GCN 5 Vega 20 · 2018-11-18used	32 GB	1,024 GB/s	—	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
AMD Ryzen AI 9 HX 370SoC / APU · Zen 5 + RDNA 3.5 "Strix Point" · 2024	32 GB unified	120 GB/s	9.0–19 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Intel Arc Pro B65Workstation GPU · Xe2 Battlemage BMG-G31 · 2026	32 GB	608 GB/s	35–72 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Intel Arc Pro B70Workstation GPU · Xe2 Battlemage BMG-G31 · 2026	32 GB	608 GB/s	35–72 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$949
Beelink SER9 (Ryzen AI 9, 32 GB)Mini PC · Zen 5 + RDNA 3.5 "Strix Point" · 2025	32 GB unified	120 GB/s	9.0–19 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$859
AMD Radeon AI PRO R9700Workstation GPU · RDNA 4 Navi 48 · 2025	32 GB	640 GB/s	42–87 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$1,299
NVIDIA RTX 5000 Ada GenerationWorkstation GPU · Ada Lovelace AD102 · 2023-08-09	32 GB	576 GB/s	56–107 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA GeForce RTX 5090Consumer GPU · Blackwell GB202 · 2025	32 GB	1,792 GB/s	113–234 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$4,700
RTX 5090 Desktop (32 GB VRAM, 64 GB RAM)AI station · Blackwell GB202 · 2025	32 GB	1,792 GB/s	113–234 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA RTX PRO 4500 Blackwell (Workstation Edition)Workstation GPU · Blackwell GB203 · 2025-03-18	32 GB	896 GB/s	70–145 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M3 ProApple Silicon · ARM, 3nm TSMC · 2023	27 GB unified	153 GB/s	9.7–20 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M4 Max (32-core GPU)Apple Silicon · ARM, 3nm TSMC · 2024	27 GB unified	410 GB/s	24–50 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M5 Max (32-core GPU)Apple Silicon · ARM, 3nm TSMC · 2026	27 GB unified	460 GB/s	27–56 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Mac Studio (M5 Max 32-core GPU, 36 GB)AI station · ARM, 3nm TSMC · 2026	27 GB unified	460 GB/s	27–56 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$2,499
MacBook Pro 14" (M5 Max 32-core GPU, 36 GB)Laptop · ARM, 3nm TSMC · 2026	27 GB unified	460 GB/s	27–56 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M1 ProApple Silicon · ARM, 5nm TSMC · 2021	24 GB unified	200 GB/s	13–26 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M2 ProApple Silicon · ARM, 5nm TSMC · 2023	24 GB unified	200 GB/s	13–26 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M4Apple Silicon · ARM, 3nm TSMC · 2024	24 GB unified	120 GB/s	7.7–16 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M5Apple Silicon · ARM, 3nm TSMC · 2025	24 GB unified	153 GB/s	9.7–20 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Apple M6Apple Silicon · ARM, 2nm TSMC · 2026	24 GB unified	170 GB/s	11–22 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Intel Arc Pro B60Workstation GPU · Xe2 Battlemage BMG-G21 · 2025	24 GB	456 GB/s	28–57 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
Mac mini (M6, 32 GB)Mini PC · ARM, 2nm TSMC · 2026	24 GB unified	170 GB/s	11–22 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
MacBook Air 13" (M5, 32 GB)Laptop · ARM, 3nm TSMC · 2026	24 GB unified	153 GB/s	9.7–20 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
MacBook Air 15" (M5, 32 GB)Laptop · ARM, 3nm TSMC · 2026	24 GB unified	153 GB/s	9.7–20 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
MacBook Pro 14" (M5, 32 GB)Laptop · ARM, 3nm TSMC · 2025	24 GB unified	153 GB/s	9.7–20 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA A10Datacenter GPU · Ampere GA102 · 2021-04-12	24 GB	600 GB/s	41–79 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA L4Datacenter GPU · Ada Lovelace AD104 · 2023-03-21	24 GB	300 GB/s	32–61 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA GeForce RTX 3090Consumer GPU · Ampere GA102 · 2020	24 GB	936 GB/s	60–115 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$1,499
RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)AI station · Ampere GA102 · 2020	24 GB	936 GB/s	60–115 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA GeForce RTX 3090 TiLegacy / used GPU · Ampere GA102 · 2022used	24 GB	1,008 GB/s	63–122 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$1,999
NVIDIA GeForce RTX 4090Consumer GPU · Ada Lovelace AD102 · 2022	24 GB	1,008 GB/s	86–165 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$1,599
RTX 4090 Desktop (24 GB VRAM, 64 GB RAM)AI station · Ada Lovelace AD102 · 2022	24 GB	1,008 GB/s	86–165 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA GeForce RTX 5090 Laptop GPULaptop GPU · Blackwell GB203 · 2025	24 GB	896 GB/s	70–145 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
RTX 5090 Laptop (24 GB VRAM, 64 GB RAM)Laptop · Blackwell GB203 · 2025	24 GB	896 GB/s	70–145 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA RTX A5000Workstation GPU · Ampere GA102 · 2021-04-12used	24 GB	768 GB/s	51–98 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA RTX PRO 4000 BlackwellWorkstation GPU · Blackwell GB203 · 2025-03-18	24 GB	672 GB/s	56–116 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
NVIDIA RTX PRO 5000 Blackwell Laptop GPULaptop GPU · Blackwell GB203 · 2025	24 GB	896 GB/s	70–145 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
AMD Radeon RX 7900 XTXConsumer GPU · RDNA 3 Navi 31 · 2022	24 GB	960 GB/s	54–112 tok/s	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	$999
NVIDIA Tesla P40Datacenter GPU · Pascal GP102 · 2016-09-13used	24 GB	346 GB/s	—	132	K2 Horizon MoVA 36B-A4B · 23.4 GB	—
AMD Radeon RX 7900 XTConsumer GPU · RDNA 3 Navi 31 · 2022	20 GB	800 GB/s	47–98 tok/s	113	Nemotron-Cascade 2 30B-A3B · 19.9 GB	$899
Apple M2Apple Silicon · ARM, 5nm TSMC · 2022	18 GB unified	100 GB/s	6.4–13 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
Apple M3Apple Silicon · ARM, 3nm TSMC · 2023	18 GB unified	100 GB/s	6.4–13 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
Apple M4 ProApple Silicon · ARM, 3nm TSMC · 2024	18 GB unified	273 GB/s	17–35 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
Mac mini (M5 Pro, 24 GB)Mini PC · ARM, 3nm TSMC · 2026	18 GB unified	307 GB/s	19–39 tok/s	102	Qwen 3.6 27B · 17.6 GB	$1,699
Mac mini (M6, 24 GB)Mini PC · ARM, 2nm TSMC · 2026	18 GB unified	170 GB/s	11–22 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
MacBook Air 13" (M5, 24 GB)Laptop · ARM, 3nm TSMC · 2026	18 GB unified	153 GB/s	9.7–20 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
MacBook Air 15" (M5, 24 GB)Laptop · ARM, 3nm TSMC · 2026	18 GB unified	153 GB/s	9.7–20 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
MacBook Pro 14" (M5, 24 GB)Laptop · ARM, 3nm TSMC · 2025	18 GB unified	153 GB/s	9.7–20 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
MacBook Pro 14" (M5 Pro, 24 GB)Laptop · ARM, 3nm TSMC · 2026	18 GB unified	307 GB/s	19–39 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
MacBook Pro 16" (M5 Pro, 24 GB)Laptop · ARM, 3nm TSMC · 2026	18 GB unified	307 GB/s	19–39 tok/s	102	Qwen 3.6 27B · 17.6 GB	—
MacBook Pro 14" (M4 Pro, 24 GB)Laptop · ARM, 3nm TSMC · 2024	18 GB unified	273 GB/s	17–35 tok/s	102	Qwen 3.6 27B · 17.6 GB	$1,999
Intel Arc Pro B50Workstation GPU · Xe2 Battlemage BMG-G21 · 2025	16 GB	224 GB/s	15–31 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$349
NVIDIA GeForce RTX 3080 Ti Laptop GPULaptop GPU · Ampere GA103 · 2022used	16 GB	512 GB/s	36–69 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
RTX 3080 Ti Laptop (16 GB VRAM, 32 GB RAM)Laptop · Ampere GA103 · 2022used	16 GB	512 GB/s	36–69 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
NVIDIA GeForce RTX 4060 Ti 16GBConsumer GPU · Ada Lovelace AD106 · 2023	16 GB	288 GB/s	31–59 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$499
NVIDIA GeForce RTX 4070 Ti SuperConsumer GPU · Ada Lovelace AD103 · 2024	16 GB	672 GB/s	63–121 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$799
NVIDIA GeForce RTX 4080Consumer GPU · Ada Lovelace AD103 · 2022	16 GB	717 GB/s	66–127 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$1,199
NVIDIA GeForce RTX 4080 SuperConsumer GPU · Ada Lovelace AD103 · 2024	16 GB	736 GB/s	68–130 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$999
NVIDIA GeForce RTX 4090 Laptop GPULaptop GPU · Ada Lovelace AD103 · 2023	16 GB	576 GB/s	56–107 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
RTX 4090 Laptop (16 GB VRAM, 32 GB RAM)Laptop · Ada Lovelace AD103 · 2023	16 GB	576 GB/s	56–107 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
NVIDIA GeForce RTX 5060 Ti 16GBConsumer GPU · Blackwell GB206 · 2025	16 GB	448 GB/s	40–83 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$805
RTX 5060 Ti 16 GB Desktop (16 GB VRAM, 32 GB RAM)AI station · Blackwell GB206 · 2025	16 GB	448 GB/s	40–83 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
NVIDIA GeForce RTX 5070 TiConsumer GPU · Blackwell GB205 · 2025	16 GB	896 GB/s	70–145 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$749
NVIDIA GeForce RTX 5080Consumer GPU · Blackwell GB203 · 2025	16 GB	960 GB/s	74–153 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$999
RTX 5080 Desktop (16 GB VRAM, 32 GB RAM)AI station · Blackwell GB203 · 2025	16 GB	960 GB/s	74–153 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
NVIDIA GeForce RTX 5080 Laptop GPULaptop GPU · Blackwell GB203 · 2025	16 GB	896 GB/s	70–145 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
RTX 5080 Laptop (16 GB VRAM, 32 GB RAM)Laptop · Blackwell GB203 · 2025	16 GB	896 GB/s	70–145 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
NVIDIA RTX PRO 2000 BlackwellWorkstation GPU · Blackwell GB206 · 2025-08-11	16 GB	288 GB/s	27–56 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
AMD Radeon RX 6800 XTLegacy / used GPU · RDNA 2 Navi 21 · 2020-11-18used	16 GB	512 GB/s	—	96	LFM2.5-24B-A2B · 15.3 GB	—
AMD Radeon RX 6900 XTLegacy / used GPU · RDNA 2 Navi 21 · 2020-12-08used	16 GB	512 GB/s	—	96	LFM2.5-24B-A2B · 15.3 GB	—
AMD Radeon RX 7600 XTConsumer GPU · RDNA 3 Navi 33 · 2024-01-24	16 GB	288 GB/s	20–42 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
AMD Radeon RX 7800 XTConsumer GPU · RDNA 3 Navi 32 · 2023	16 GB	624 GB/s	39–81 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$499
AMD Radeon RX 7900 GREConsumer GPU · RDNA 3 Navi 31 · 2023-07-27	16 GB	576 GB/s	36–76 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	—
AMD Radeon RX 9060 XT 16GBConsumer GPU · RDNA 4 Navi 44 · 2025	16 GB	320 GB/s	24–49 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$349
AMD Radeon RX 9070Consumer GPU · RDNA 4 Navi 48 · 2025	16 GB	640 GB/s	42–87 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$549
AMD Radeon RX 9070 XTConsumer GPU · RDNA 4 Navi 48 · 2025	16 GB	640 GB/s	42–87 tok/s	96	LFM2.5-24B-A2B · 15.3 GB	$599
Apple M1Apple Silicon · ARM, 5nm TSMC · 2020	12 GB unified	68 GB/s	4.4–9.2 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
Intel Arc B580Consumer GPU · Xe2 Battlemage BMG-G21 · 2024	12 GB	456 GB/s	28–57 tok/s	85	Cosmos 3 Nano · 10.5 GB	$249
Mac mini (M4, 16 GB)Mini PC · ARM, 3nm TSMC · 2024	12 GB unified	120 GB/s	7.7–16 tok/s	85	Cosmos 3 Nano · 10.5 GB	$599
Mac mini (M6, 16 GB)Mini PC · ARM, 2nm TSMC · 2026	12 GB unified	170 GB/s	11–22 tok/s	85	Cosmos 3 Nano · 10.5 GB	$899
MacBook Air 13" (M5, 16 GB)Laptop · ARM, 3nm TSMC · 2026	12 GB unified	153 GB/s	9.7–20 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
MacBook Air 15" (M5, 16 GB)Laptop · ARM, 3nm TSMC · 2026	12 GB unified	153 GB/s	9.7–20 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
MacBook Air (M4, 16 GB)Laptop · ARM, 3nm TSMC · 2024	12 GB unified	120 GB/s	7.7–16 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
MacBook Pro 14" (M5, 16 GB)Laptop · ARM, 3nm TSMC · 2025	12 GB unified	153 GB/s	9.7–20 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
NVIDIA GeForce RTX 3060 (12GB)Consumer GPU · Ampere GA106 · 2021	12 GB	360 GB/s	26–50 tok/s	85	Cosmos 3 Nano · 10.5 GB	$329
RTX 3060 12 GB Desktop (12 GB VRAM, 32 GB RAM)AI station · Ampere GA106 · 2021	12 GB	360 GB/s	26–50 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
NVIDIA GeForce RTX 3080 12GBLegacy / used GPU · Ampere GA102 · 2022used	12 GB	912 GB/s	59–112 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
NVIDIA GeForce RTX 3080 TiLegacy / used GPU · Ampere GA102 · 2021used	12 GB	912 GB/s	59–112 tok/s	85	Cosmos 3 Nano · 10.5 GB	$1,199
NVIDIA GeForce RTX 4070Consumer GPU · Ada Lovelace AD104 · 2023	12 GB	504 GB/s	50–96 tok/s	85	Cosmos 3 Nano · 10.5 GB	$599
NVIDIA GeForce RTX 4070 SuperConsumer GPU · Ada Lovelace AD104 · 2024	12 GB	504 GB/s	50–96 tok/s	85	Cosmos 3 Nano · 10.5 GB	$599
NVIDIA GeForce RTX 4070 TiConsumer GPU · Ada Lovelace AD104 · 2023	12 GB	504 GB/s	50–96 tok/s	85	Cosmos 3 Nano · 10.5 GB	$799
NVIDIA GeForce RTX 4080 Laptop GPULaptop GPU · Ada Lovelace AD104 · 2023	12 GB	432 GB/s	44–84 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
RTX 4080 Laptop (12 GB VRAM, 32 GB RAM)Laptop · Ada Lovelace AD104 · 2023	12 GB	432 GB/s	44–84 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
NVIDIA GeForce RTX 5070Consumer GPU · Blackwell GB205 · 2025	12 GB	672 GB/s	56–116 tok/s	85	Cosmos 3 Nano · 10.5 GB	$549
NVIDIA GeForce RTX 5070 Ti Laptop GPULaptop GPU · Blackwell GB205 · 2025	12 GB	672 GB/s	56–116 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
RTX 5070 Ti Laptop (12 GB VRAM, 32 GB RAM)Laptop · Blackwell GB205 · 2025	12 GB	672 GB/s	56–116 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
AMD Radeon RX 6700 XTLegacy / used GPU · RDNA 2 Navi 22 · 2021-03-01used	12 GB	384 GB/s	—	85	Cosmos 3 Nano · 10.5 GB	—
AMD Radeon RX 7700 XTConsumer GPU · RDNA 3 Navi 32 · 2023	12 GB	432 GB/s	29–60 tok/s	85	Cosmos 3 Nano · 10.5 GB	$449
AMD Radeon RX 9070 GREConsumer GPU · RDNA 4 Navi 48 · 2025-05-08	12 GB	432 GB/s	31–64 tok/s	85	Cosmos 3 Nano · 10.5 GB	—
NVIDIA GeForce GTX 1080 TiLegacy / used GPU · Pascal GP102 · 2017-03-10used	11 GB	484 GB/s	—	85	Cosmos 3 Nano · 10.5 GB	$699
NVIDIA GeForce RTX 2080 TiLegacy / used GPU · Turing TU102 · 2018-09-27used	11 GB	616 GB/s	—	85	Cosmos 3 Nano · 10.5 GB	$999
Intel Arc B570Consumer GPU · Xe2 Battlemage BMG-G21 · 2025	10 GB	380 GB/s	24–49 tok/s	83	Qwen 3 14B · 9.7 GB	$219
NVIDIA GeForce RTX 3080 (10GB)Consumer GPU · Ampere GA102 · 2020	10 GB	760 GB/s	50–97 tok/s	83	Qwen 3 14B · 9.7 GB	$699
NVIDIA GeForce RTX 3060 TiLegacy / used GPU · Ampere GA104 · 2020-12-02used	8 GB	448 GB/s	32–61 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$399
NVIDIA GeForce RTX 3070Consumer GPU · Ampere GA104 · 2020	8 GB	448 GB/s	32–61 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$499
NVIDIA GeForce RTX 3070 TiConsumer GPU · Ampere GA104 · 2021	8 GB	608 GB/s	42–80 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$599
NVIDIA GeForce RTX 4060Consumer GPU · Ada Lovelace AD107 · 2023	8 GB	272 GB/s	29–56 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$299
NVIDIA GeForce RTX 4060 Laptop GPULaptop GPU · Ada Lovelace AD107 · 2023	8 GB	256 GB/s	28–53 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	—
RTX 4060 Laptop (8 GB VRAM, 16 GB RAM)Laptop · Ada Lovelace AD107 · 2023	8 GB	256 GB/s	28–53 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$1,099
NVIDIA GeForce RTX 4060 Ti 8GBConsumer GPU · Ada Lovelace AD106 · 2023	8 GB	288 GB/s	31–59 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$399
NVIDIA GeForce RTX 5050Consumer GPU · Blackwell GB207 · 2025-07-01	8 GB	320 GB/s	30–61 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$249
NVIDIA GeForce RTX 5060Consumer GPU · Blackwell GB206 · 2025	8 GB	448 GB/s	40–83 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$299
NVIDIA GeForce RTX 5060 Laptop GPULaptop GPU · Blackwell GB206 · 2025	8 GB	384 GB/s	35–72 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	—
RTX 5060 Laptop (8 GB VRAM, 16 GB RAM)Laptop · Blackwell GB206 · 2025	8 GB	384 GB/s	35–72 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	—
NVIDIA GeForce RTX 5060 Ti 8GBConsumer GPU · Blackwell GB206 · 2025	8 GB	448 GB/s	40–83 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$379
NVIDIA GeForce RTX 5070 Laptop GPULaptop GPU · Blackwell GB206 · 2025	8 GB	384 GB/s	35–72 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	—
RTX 5070 Laptop (8 GB VRAM, 16 GB RAM)Laptop · Blackwell GB206 · 2025	8 GB	384 GB/s	35–72 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	—
AMD Radeon RX 7600Consumer GPU · RDNA 3 Navi 33 · 2023-05-24	8 GB	288 GB/s	20–42 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	—
AMD Radeon RX 9060 XT 8GBConsumer GPU · RDNA 4 Navi 44 · 2025	8 GB	320 GB/s	24–49 tok/s	72	Bielik PL 11B v3.0 Instruct · 7.4 GB	$299

Browse by memory

8 GB — models that fit 12 GB — models that fit 16 GB — models that fit 24 GB — models that fit 32 GB — models that fit 48 GB — models that fit 64 GB — models that fit 80 GB — models that fit

Speed is Llama 3.1 8B at Q4_K_M: a bandwidth-roofline estimate shown as a range unless the device page marks it measured. Mac memory is the full unified pool; about 75% of it is usable by a model. Fit counts catalogued model variants whose Q4_K_M weights plus runtime overhead fit the usable memory, before any context is added. Methodology

Browse

Tools

Guides

Browse by type

Browse by memory

Keep going