AIベンチマーク

Neural Engine / AI指数。

1.5.0~1.7.0 / iOS 26.3+, iPadOS 26.3+, macOS 26.3+ / / 2026.04.16

iPhone

CPU FP32GPU FP32ANE FP32CPU FP16GPU FP16ANE FP16CPU INT8GPU INT8ANE INT8

(2P+4E+4GPU+8NE) A13 Bionic
1845
912
1235
3309
1161
3598
2755
517
739
(2P+4E+4GPU+16NE) A14 Bionic
2906
2179
2919
5073
2444
13305
4231
2305
13456
(2P+4E+4GPU+16NE) A15 Bionic
3356
3392
3395
6100
3765
20809
4941
3613
23412
(2P+4E+5GPU+16NE) A15 Bionic
3386
4373
3399
6162
4952
20600
4987
4671
23619
(2P+4E+5GPU+16NE) A16 Bionic
3667
4615
3680
6660
5082
23525
5423
4770
27494
(2P+4E+6GPU+16NE) A17 Pro
4115
5570
4089
7387
6612
28802
6042
6436
40262
(2P+4E+5GPU+16NE) A18
4745
6367
4708
8005
6483
35878
6456
6165
49251
(2P+4E+6GPU+16NE) A18 Pro
4935
7313
4948
8419
8634
35559
6702
8198
49132
(2P+4E+5GPU+16NE) A19
4947
8793
4974
8116
17698
37946
6601
18511
52617
(2P+4E+5GPU+16NE) A19 Pro
5068
7309
5061
8313
15241
34990
6528
14287
48251
(2P+4E+6GPU+16NE) A19 Pro
5479
10134
5519
8718
19270
39013
6923
20706
54233

iPad

CPU FP32GPU FP32ANE FP32CPU FP16GPU FP16ANE FP16CPU INT8GPU INT8ANE INT8

(2P+4E+4GPU+8NE) A12 Bionic
727
613
541
1200
724
2090
1078
254
381
(4P+4E+7GPU+8NE) A12X Bionic
1069
1181
799
1668
1430
2548
1430
420
475
(4P+4E+8GPU+8NE) A12Z Bionic
1079
1313
805
1665
1616
2564
1438
444
476
(2P+4E+4GPU+8NE) A13 Bionic
1842
862
1240
3298
1216
3500
2742
514
741
(2P+4E+4GPU+16NE) A14 Bionic
2866
2953
2879
5038
3388
13092
4179
3291
13453
(2P+4E+5GPU+16NE) A15 Bionic
3170
3293
3218
5713
3564
20086
4641
3459
22441
(2P+3E+4GPU+16NE) A16 Bionic
3695
3931
3711
6674
4366
23596
5409
3738
27185
(2P+4E+5GPU+16NE) A17 Pro
4084
4872
4154
7422
6107
28785
6027
5303
40242
(4P+4E+8GPU+16NE) M1
3298
5994
3291
5801
6682
15781
4712
6378
15205
(4P+4E+10GPU+16NE) M2
3520
8694
3529
6387
9778
28675
5178
8828
32376
(4P+4E+9GPU+16NE) M3
4315
9652
4280
7556
11045
33098
6101
9752
37492
(4P+6E+10GPU+16NE) M4
5224
11112
5209
8609
13062
39747
6867
12407
56784
(4P+6E+10GPU+16NE) M5
5680
14519
5642
9193
25820
42982
7041
27295
61575

Mac

CPU FP32GPU FP32ANE FP32CPU FP16GPU FP16ANE FP16CPU INT8GPU INT8ANE INT8

(2P+4E+5GPU+16NE) A18 Pro
5237
7642
5178
8689
8847
35654
6903
8224
49691
(4P+4E+8GPU+16NE) M1
3278
6299
3266
5750
7328
16404
4677
6860
16135
(8P+2E+16GPU+16NE) M1 Pro
3916
10004
3912
6484
10873
17942
5170
9836
17817
(8P+2E+32GPU+16NE) M1 Max
3945
14439
3975
6447
16012
16596
5170
15041
16442
(16P+4E+64GPU+32NE) M1 Ultra
3785
15429
4012
6093
16737
15637
4913
15621
15318
(4P+4E+10GPU+16NE) M2
3516
8701
3524
6396
9833
28895
5210
9213
32642
(8P+4E+19GPU+16NE) M2 Pro
4350
13492
4362
7274
14644
29729
5823
13662
33548
(8P+4E+38GPU+16NE) M2 Max
4379
18134
4385
7337
20058
29681
5891
18878
33558
(16P+8E+76GPU+32NE) M2 Ultra
4815
21183
4830
7755
22810
28946
6159
21354
32552
(4P+4E+10GPU+16NE) M3
4286
10189
4275
7458
11846
33117
6060
10693
37671
(6P+6E+18GPU+16NE) M3 Pro
4302
14830
4315
7539
17460
32619
6131
16234
37112
(12P+4E+40GPU+16NE) M3 Max
4970
22355
5056
8056
25121
32628
6461
23644
36314
(24P+8E+80GPU+32NE) M3 Ultra
5781
24216
5796
9092
26522
31904
7164
24762
35097
(4P+6E+10GPU+16NE) M4
5172
11864
5163
8556
13996
39975
6864
13208
57148
(10P+4E+20GPU+16NE) M4 Pro
6096
18567
6075
10081
20477
42361
7780
19643
59035
(12P+4E+40GPU+16NE) M4 Max
6297
26592
6273
10273
29525
41741
7972
27356
57980
(4P+6E+10GPU+16NE) M5
5638
14837
5614
9207
27932
43429
7102
28253
61570
(6P+12M+40GPU+16NE) M5 Max
5526
29933
5487
9575
47708
49142
7435
44140
67005
(6P+12M+20GPU+16NE) M5 Pro
5449
21793
5409
9523
37936
49482
7422
36590
67719

LLaMa 7B ( ：tok/s) / llama.cpp / Backend Metal / 2025.11.27

iPhone

INT4 (Q4_0 TG)INT8 (Q8_0 TG)FP16 (TG)

(51.2GB/s, 6GPU) A17 Pro
9.01
-
-
(60.0GB/s, 6GPU) A18 Pro
11.64
-
-
(68.26GB/s, 5GPU) A19 Pro
12.74
-
-
(68.26GB/s ,7GPU) M1
14.19
7.92
-
(68.26GB/s, 8GPU) M1
14.15
7.91
-
(204.8GB/s, 14GPU) M1 Pro
35.52
21.95
12.75
(204.8GB/s, 16GPU) M1 Pro
36.41
22.34
12.75
(409.6GB/s, 24GPU) M1 Max
54.61
37.81
22.55
(409.6GB/s, 32GPU) M1 Max
61.19
40.2
23.03
(819.2GB/s, 48GPU) M1 Ultra
74.93
55.69
33.92
(819.2GB/s, 64GPU) M1 Ultra
83.73
59.87
37.01
(102.4GB/s, 8GPU) M2
21.7
12.18
-
(102.4GB/s, 10GPU) M2
21.91
12.21
6.72
(204.8GB/s, 16GPU) M2 Pro
37.87
22.7
12.47
(204.8GB/s, 19GPU) M2 Pro
38.86
23.01
13.06
(409.6GB/s, 30GPU) M2 Max
60.99
39.97
24.16
(409.6GB/s, 38GPU) M2 Max
65.95
41.83
24.65
(819.2GB/s, 60GPU) M2 Ultra
88.64
62.14
39.86
(819.2GB/s, 76GPU) M2 Ultra
94.27
66.64
41.02
(102.4GB/s, 10GPU) M3
21.34
12.27
-
(153.6GB/s, 14GPU) M3 Pro
30.65
17.44
-
(153.6GB/s, 18GPU) M3 Pro
30.74
17.53
9.89
(307.2GB/s, 30GPU) M3 Max
56.58
34.3
19.54
(409.6GB/s, 40GPU) M3 Max
66.31
42.75
25.09
(819.2GB/s, 60GPU) M3 Ultra
88.4
63.55
42.24
(819.2GB/s, 80GPU) M3 Ultra
92.14
64.16
40.1
(120.0GB/s, 10GPU) M4
24.11
13.54
7.43
(273.0GB/s, 16GPU) M4 Pro
49.64
30.54
17.19
(273.0GB/s, 20GPU) M4 Pro
50.74
30.69
17.18
(546.0GB/s, 40GPU) M4 Max
83.06
54.05
31.64
(410.0GB/s, 32GPU) M4 Max
69.24
43.8
25.58
( , 153.6GB/s, 10GPU) M5
29.62
16.62
9.43
( , 153.6GB/s, 10GPU) M5
31.02
17.5
9.67

TinyLlama 1.1B ( ：tok/s) / llama.cpp / Backend Metal / 2025.11.27

iPad

INT4 (Q4_0 TG)INT8 (Q8_0 TG)FP16 (TG)

(34.13GB/s, 4GPU) A14 Bionic
39.21
24.11
10.26
(34.13GB/s, 4GPU) A15 Bionic
39.03
24.12
-
(34.13GB/s, 5GPU) A15 Bionic
39.09
23.84
13.66
(51.2GB/s, 5GPU) A16 Bionic
54.24
34.3
20.06
(51.2GB/s, 6GPU) A17 Pro
56.86
35.6
20.23
(60.0GB/s, 6GPU) A18 Pro
69.77
42.11
23.84
(120.0GB/s, 10GPU) M4
121.2
75.33
46.21
(68.26GB/s, 5GPU) A19 Pro
87.33
53.92
29.98

Phi-2 2.7B ( ：tok/s) / llama.cpp / Backend Metal / 2025.11.27

Mac

INT4 (Q4_0 TG)INT8 (Q8_0 TG)

(34.13GB/s, 4GPU) A14 Bionic
8.52
-
(34.13GB/s, 5GPU) A15 Bionic
16.73
-
(51.2GB/s, 5GPU) A16 Bionic
23.31
14.06
(51.2GB/s, 6GPU) A17 Pro
24.71
14.74
(60.0GB/s, 6GPU) A18 Pro
-
16.93
(68.26GB/s, 5GPU) A19 Pro
36.01
21.42