← Full benchmark database
Value leaderboard

Performance, cost, power and VRAM — weighted by what you actually value

Four axes for Qwen 3 8B, each normalised across the hardware shown, so you can see at a glance where a machine is strong and where it gives something up.

There is no score here until you make one. A single blended number with weights we picked would look objective while hiding an opinion you never saw — so the default is the four axes, and the ranking appears once you say what matters to you. Your weights go in the URL, so a link you share shows what you valued.

Machines that run it108Qwen 3 8B at Q4_K_M, 4K context
Fastest238 tok/sNVIDIA H100 80GB (SXM)
Most efficient1.50 tok/WApple M5 Max · SoC power
What this ranking rests on: 1 measured of 108 rows.Three minutes with llama-bench moves your hardware out of the estimate column →
1 measured17 modelled on a constant fitted from real runs90 on a constant nobody has calibrated
Q4_K_M · 4K context · decode speed

What do you actually want from this machine?

Pick a starting point — each is an opinion, named for the person it suits — or set the sliders yourself. Nothing is ranked until you do.

HardwareSpeedRunning costEfficiencyMemoryEvidence
1AMD Ryzen AI 9 HX 370
0.03
0.40
0.07
0.05
2AMD Ryzen AI Max+ 395
0.09
0.33
0.05
0.24
3AMD Ryzen AI Max+ PRO 495
0.10
0.00
0.00
0.37
4Apple M1
0.00
0.58
0.13
0.02
5Apple M1 Max
0.12
0.97
0.78
0.11
6Apple M1 Pro
0.05
0.83
0.34
0.05
7Apple M1 Ultra
0.25
0.95
0.70
0.24
8Apple M2
0.01
0.75
0.24
0.03
9Apple M2 Max
0.12
0.92
0.54
0.17
10Apple M2 Pro
0.05
0.83
0.34
0.05
11Apple M2 Ultra
0.25
0.95
0.70
0.37
12Apple M3
0.01
0.71
0.21
0.03
13Apple M3 Max
0.12
0.95
0.63
0.24
14Apple M3 Max (30-core GPU)
0.09
0.89
0.46
0.17
15Apple M3 Pro
0.03
0.75
0.24
0.06
16Apple M3 Ultra
0.25
0.96
0.70
1.00
17Apple M4
0.02
0.78
0.27
0.05
18Apple M4 Max
0.17
0.99
0.85
0.24
19Apple M4 Max (32-core GPU)
0.13
0.95
0.63
0.06
20Apple M4 Pro
0.08
0.91
0.50
0.03
21Apple M5
0.03
0.84
0.35
0.05
22Apple M5 Max
0.19
1.00
1.00
0.24
23Apple M5 Max (32-core GPU)
0.14
0.97
0.70
0.06
24Apple M5 Pro
0.09
0.93
0.57
0.11
25Apple M5 Ultra
0.36
0.88
0.43
1.00
26Apple M6
0.04
0.84
0.36
0.05
27Apple M6 (16 GB, 153 GB/s)
0.03
0.00
0.00
0.02
28Intel Arc B570
0.12
0.36
0.06
0.00
29Intel Arc B580
0.15
0.30
0.04
0.01
30Intel Arc Pro B65
0.20
0.44
0.07
0.05
31Intel Arc Pro B70
0.20
0.35
0.05
0.05
32Intel Arc Pro B50
0.07
0.56
0.12
0.02
33Intel Arc Pro B60
0.15
0.26
0.04
0.03
34NVIDIA A10
0.22
0.66
0.17
0.03
35NVIDIA A100 80GB (PCIe)
0.60
0.76
0.25
0.14
36NVIDIA DGX Spark
0.14
0.42
0.07
0.24
37NVIDIA GB10 Grace Blackwell
0.14
0.47
0.08
0.24
38NVIDIA H100 80GB (PCIe)
0.75
0.78
0.27
0.14
39NVIDIA H100 80GB (SXM)
1.00
0.60
0.13
0.14
40NVIDIA Jetson Thor (T5000)
0.14
0.52
0.10
0.24
41NVIDIA L4
0.17
0.84
0.35
0.03
42NVIDIA L40
0.44
0.63
0.15
0.08
43NVIDIA L40S
0.44
0.55
0.11
0.08
44AMD Radeon AI PRO R9600D
0.24
0.70
0.19
0.05
45AMD Radeon AI PRO R9700
0.24
0.27
0.04
0.05
46AMD Radeon AI PRO R9700S
0.24
0.00
0.00
0.05
47AMD Radeon PRO W7900
0.29
0.42
0.07
0.08
48NVIDIA GeForce RTX 3060 (12GB)
0.13
0.30
0.04
0.01
49NVIDIA GeForce RTX 3060 Ti
0.17
0.33
0.05
0.00
50NVIDIA GeForce RTX 3070
0.17
0.25
0.04
0.00
51NVIDIA GeForce RTX 3070 Ti
0.23
0.25
0.03
0.00
52NVIDIA GeForce RTX 3080 (10GB)
0.28
0.32
0.04
0.00
53NVIDIA GeForce RTX 3080 12GB
0.33
0.37
0.06
0.01
54NVIDIA GeForce RTX 3080 Ti
0.33
0.37
0.06
0.01
55NVIDIA GeForce RTX 3080 Ti Laptop GPU
0.19
0.00
0.00
0.02
56NVIDIA GeForce RTX 3090
0.34
0.39
0.06
0.03
57NVIDIA GeForce RTX 3090 Ti
0.36
0.23
0.03
0.03
58NVIDIA GeForce RTX 4060
0.15
0.62
0.15
0.00
59NVIDIA GeForce RTX 4060 Laptop GPU
0.14
0.60
0.13
0.00
60NVIDIA GeForce RTX 4060 Ti 16GB
0.16
0.44
0.07
0.02
61NVIDIA GeForce RTX 4060 Ti 8GB
0.16
0.47
0.08
0.00
62NVIDIA GeForce RTX 4070
0.28
0.62
0.14
0.01
63NVIDIA GeForce RTX 4070 Super
0.28
0.57
0.12
0.01
64NVIDIA GeForce RTX 4070 Ti
0.28
0.40
0.07
0.01
65NVIDIA GeForce RTX 4070 Ti Super
0.36
0.56
0.11
0.02
66NVIDIA GeForce RTX 4080
0.38
0.52
0.10
0.02
67NVIDIA GeForce RTX 4080 Laptop GPU
0.24
0.00
0.00
0.01
68NVIDIA GeForce RTX 4080 Super
0.39
0.53
0.10
0.02
69NVIDIA GeForce RTX 4090
0.58
0.55
0.11
0.03
70NVIDIA GeForce RTX 4090 Laptop GPU
0.31
0.78
0.28
0.02
71NVIDIA RTX 5000 Ada Generation
0.31
0.56
0.12
0.05
72NVIDIA GeForce RTX 5050
0.16
0.60
0.13
0.00
73NVIDIA GeForce RTX 5060
0.23
0.69
0.19
0.00
74NVIDIA GeForce RTX 5060 Laptop GPU
0.20
0.00
0.00
0.00
75NVIDIA GeForce RTX 5060 Ti 16GB
0.23
0.58
0.13
0.02
76NVIDIA GeForce RTX 5060 Ti 8GB
0.23
0.58
0.13
0.00
77NVIDIA GeForce RTX 5070
0.33
0.59
0.13
0.01
78NVIDIA GeForce RTX 5070 Laptop GPU
0.20
0.78
0.27
0.00
79NVIDIA GeForce RTX 5070 Ti
0.42
0.61
0.14
0.02
80NVIDIA GeForce RTX 5070 Ti Laptop GPU
0.33
0.88
0.43
0.01
81NVIDIA GeForce RTX 5080
0.45
0.54
0.11
0.02
82NVIDIA GeForce RTX 5080 Laptop GPU
0.42
0.87
0.40
0.02
83NVIDIA GeForce RTX 5090
0.70
0.52
0.10
0.05
84NVIDIA GeForce RTX 5090 Laptop GPU
0.42
0.87
0.40
0.03
85NVIDIA RTX 6000 Ada Generation
0.48
0.67
0.17
0.08
86NVIDIA RTX A5000
0.28
0.55
0.11
0.03
87NVIDIA RTX A6000
0.28
0.38
0.06
0.08
88NVIDIA RTX PRO 2000 Blackwell
0.14
0.81
0.31
0.02
89NVIDIA RTX PRO 4000 Blackwell
0.33
0.00
0.00
0.03
90NVIDIA RTX PRO 4000 Blackwell SFF
0.22
0.91
0.49
0.03
91NVIDIA RTX PRO 4500 Blackwell (Workstation Edition)
0.42
0.78
0.27
0.05
92NVIDIA RTX PRO 5000 Blackwell (48 GB)
0.58
0.74
0.23
0.08
93NVIDIA RTX PRO 5000 Blackwell (72 GB)
0.58
0.74
0.23
0.13
94NVIDIA RTX PRO 5000 Blackwell Laptop GPU
0.42
0.00
0.00
0.03
95NVIDIA RTX PRO 6000 Blackwell
0.70
0.49
0.09
0.17
96NVIDIA RTX PRO 6000 Blackwell (Max-Q Edition)
0.70
0.00
0.00
0.17
97AMD Radeon RX 7600
0.10
0.15
0.01
0.00
98AMD Radeon RX 7600 XT
0.10
0.00
0.00
0.02
99AMD Radeon RX 7700 XT
0.16
0.11
0.01
0.01
100AMD Radeon RX 7800 XT
0.22
0.32
0.04
0.02
101AMD Radeon RX 7900 GRE
0.21
0.27
0.04
0.02
102AMD Radeon RX 7900 XT
0.28
0.33
0.04
0.02
103AMD Radeon RX 7900 XTX
0.32
0.34
0.05
0.03
104AMD Radeon RX 9060 XT 16GB
0.12
0.32
0.04
0.02
105AMD Radeon RX 9060 XT 8GB
0.12
0.37
0.06
0.00
106AMD Radeon RX 9070
0.24
0.50
0.10
0.02
107AMD Radeon RX 9070 GRE
0.17
0.27
0.04
0.01
108AMD Radeon RX 9070 XT
0.24
0.26
0.04
0.02

SoC marks the 30 unified-memory systems in this ranking. Their power figure is whole-package power — CPU, GPU and memory controller on one die — where a discrete card reports board power for the card alone. Neither is the wall draw of the complete machine. They are not the same measurement, so read the ranking as a real difference in efficiency that is partly a difference in what is being counted.

How this is calculated

Decode speed comes from the same engine as the rest of the site — a three-term latency model with efficiency fitted per GPU architecture. Full methodology

each axis is min-max normalised across the hardware shown:
  axis = (value - min) / (max - min)          0 = weakest, 1 = strongest

cost is inverted first, so higher is better on all four axes.
an axis every row agrees on scores 0 -- it carries no information,
and scoring it 1 would let a flat axis dominate your composite.

composite = SUM(axis x your_weight) / SUM(your_weights)

Watts is the heat, and the heat is the noise — but the mapping between them is the cooler's job, and we hold no cooler data. A 450 W card in a well-designed case can be quieter than a 250 W card with a bad fan curve, so nothing here should be read as a loudness ranking. If you own a sound meter, we would like your reading.

购买此硬件 Apple MacBook Pro M5 Max — 128 GB unified memory · 614 GB/s立即云端部署 RunPod 上的 RTX 4090 — 低至 $0.34/小时 · 价格核实于 2026-07

或在 Vast.ai 比较，低至 $0.35/小时 (typical low · varies)

作为亚马逊联盟成员，我们从符合条件的购买中获得收入。云 GPU 链接为推荐链接——我们可能获得佣金，您无需额外付费。

Tokens per wattCost per 1M tokensThe full databaseWhat running LLMs 24/7 costs