← All Mac Studio models

Mac Studio M3 Ultra

The M3 Ultra Mac Studio runs local models at 819 GB/s of memory bandwidth with 96 to 512 GB of unified memory. On Apple Silicon that memory is shared with the GPU, so the whole pool is available for weights: at 512 GB you can hold roughly a 696B dense model at Q4. Two Max dies fused together: the widest memory bus and the highest capacity Apple sells. This is the tier that runs frontier-size open weights locally.

Apple no longer sells this configuration new. It stays fully evaluated here because the used market is where most of its local AI value now sits.

By , Founder & Lead Engineer— Updated

Specifications

ChipApple M3 Ultra
CPU cores28 or 32
GPU cores60 or 80
Unified memory96, 256, or 512 GB
Memory bandwidth819 GB/s
Neural Engine36 TOPS
Released2025
AvailabilityUsed market

Memory bandwidth is faster than 90% of the Apple Silicon chips shipped in a Mac, against a 1228 GB/s peak.

Its memory ceiling is above 90% of them, against a 512 GB peak.

Pick your configuration

Every option Apple sells with this chip. The model list below recomputes against the one you pick.

GPU cores

Unified memory

Apple couples memory to the core count on this chip, so the options change with the bin above.

What each memory option runs

Unified memory is the ceiling and it is soldered, so this is the decision you cannot revisit.

96 GB unified memory

84 GB usable for weights · 819 GB/s

6,384 of 6,563 models fit, and 6,271 of them run with headroom rather than as a squeeze.

Largest model at Q4
LING 3.0 FLASH ABLITERATED · 127.49B
Best all-round pick
Ornith 1.0 35B · Q8_0 · ~158 tok/s

256 GB unified memory

225 GB usable for weights · 819 GB/s

6,493 of 6,563 models fit, and 6,424 of them run with headroom rather than as a squeeze.

Largest model at Q4
DeepSeek V3.2 REAP 345B A37B · 344.89B
Best all-round pick
Ornith 1.0 35B · Q8_0 · ~158 tok/s

512 GB unified memory

450 GB usable for weights · 819 GB/s

6,549 of 6,563 models fit, and 6,490 of them run with headroom rather than as a squeeze.

Largest model at Q4
A.X K2 · 691.69B
Best all-round pick
Ornith 1.0 35B · Q8_0 · ~158 tok/s

What a 96 GB M3 Ultra Mac Studio can run

Every model in the database against this exact configuration, at 819 GB/s. Ratings and speeds are the same numbers the model pages show.

Showing 6563 of 6563 models

General · deepreinforce-ai · 2026-06-21

Q8_0Excellent
38.6 GB40% of RAMBenchmark needed34.13B params
Run with ToolPiper

General · radixark · 2026-07-27

Q8_0Excellent
3.0 GB3% of RAM~191 tok/sEstimated2.25B params
Run with ToolPiper

General · bahushruth · 2026-06-11

Q8_0Excellent
39.2 GB41% of RAMBenchmark needed34.66B params
Run with ToolPiper

General · openbmb · 2026-05-21

Q8_0Excellent
1.7 GB2% of RAM~397 tok/sEstimated1.08B params
Run with ToolPiper

General · farbodtavakkoli · 2026-06-17

Q8_0Excellent
10.8 GB11% of RAMBenchmark needed9.23B params
Run with ToolPiper

General · internscience · 2026-07-13

Q8_0Excellent
5.6 GB6% of RAM~94 tok/sEstimated4.54B params
Run with ToolPiper

General · Liquid AI · 2026-05-28

Q8_0Excellent
9.9 GB10% of RAMBenchmark needed8.47B params
Run with ToolPiper

General · flywheel-ai · 2026-06-21

Q8_0Excellent
39.2 GB41% of RAMBenchmark needed34.66B params
Run with ToolPiper

General · weiboai · 2026-06-12

Q8_0Excellent
3.9 GB4% of RAM~139 tok/sEstimated3.09B params
Run with ToolPiper

General · Liquid AI · 2026-07-28

Q8_0Excellent
3.5 GB4% of RAM~159 tok/sEstimated2.7B params
Run with ToolPiper

General · Alibaba · 2026-06-22

Q8_0Excellent
39.2 GB41% of RAMBenchmark needed34.66B params
Run with ToolPiper

General · poolside · 2026-06-20

Q8_0Excellent
37.8 GB39% of RAMBenchmark needed33.44B params
Run with ToolPiper

General · Liquid AI · 2026-06-24

Q8_0Excellent
0.8 GB1% of RAM~1,865 tok/sEstimated0.23B params
Run with ToolPiper

General · nanbeige · 2026-07-21

Q8_0Excellent
5.2 GB5% of RAM~103 tok/sEstimated4.17B params
Run with ToolPiper

General · ma7ee7 · 2026-07-30

Q8_0Excellent
4.0 GB4% of RAM~137 tok/sEstimated3.13B params
Run with ToolPiper

General · apodex · 2026-06-07

Q8_0Excellent
40.6 GB42% of RAMBenchmark needed35.95B params
Run with ToolPiper

General · internscience · 2026-06-22

Q8_0Excellent
39.7 GB41% of RAMBenchmark needed35.11B params
Run with ToolPiper

Coding · coherelabs · 2026-06-05

Q8_0Excellent
34.5 GB36% of RAMBenchmark needed30.48B params
Run with ToolPiper

General · goekdeniz-guelmez · 2026-07-31

Q8_0Excellent
3.0 GB3% of RAM~189 tok/sEstimated2.27B params
Run with ToolPiper

General · ktruestory · 2026-05-28

Q8_0Excellent
1.7 GB2% of RAM~397 tok/sEstimated1.08B params
Run with ToolPiper

General · petrouil · 2026-07-23

Q8_0Excellent
3.4 GB4% of RAMBenchmark needed2.61B params
Run with ToolPiper

Multimodal · Alibaba · 2026-04-15

Q8_0Excellent
40.6 GB42% of RAMBenchmark needed35.95B params
Run with ToolPiper

General · deepreinforce-ai · 2026-06-21

Q8_0Excellent
9.7 GB10% of RAM~52 tok/sEstimated8.21B params
Run with ToolPiper

General · ibm-granite · 2026-04-06

Q8_0Excellent
4.3 GB4% of RAM~126 tok/sEstimated3.4B params
Run with ToolPiper

Multimodal · ibm-granite · 2026-04-16

Q8_0Excellent
5.0 GB5% of RAM~107 tok/sEstimated4B params
Run with ToolPiper

General · lukebailey181pub · 2026-04-21

Q8_0Excellent
8.2 GB9% of RAM~62 tok/sEstimated6.91B params
Run with ToolPiper

General · trevorjs · 2026-04-03

Q8_0Excellent
29.3 GB31% of RAMBenchmark needed25.81B params
Run with ToolPiper

General · empero-ai · 2026-06-19

Q8_0Excellent
11.0 GB11% of RAM~46 tok/sEstimated9.41B params
Run with ToolPiper

General · poolside · 2026-04-23

Q8_0Excellent
37.8 GB39% of RAMBenchmark needed33.44B params
Run with ToolPiper

General · huihui-ai · 2026-04-21

Q8_0Excellent
40.6 GB42% of RAMBenchmark needed35.95B params
Run with ToolPiper

Multimodal · huihui-ai · 2026-04-18

Q8_0Excellent
40.6 GB42% of RAMBenchmark needed35.95B params
Run with ToolPiper

General · goekdeniz-guelmez · 2026-07-31

Q8_0Excellent
11.3 GB12% of RAM~44 tok/sEstimated9.65B params
Run with ToolPiper

Multimodal · Google · 2026-03-11

Q8_0Excellent
30.1 GB31% of RAMBenchmark needed26.54B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-27

Q8_0Excellent
5.7 GB6% of RAM~92 tok/sEstimated4.66B params
Run with ToolPiper

Multimodal · Google · 2026-03-02

Q8_0Excellent
9.4 GB10% of RAM~54 tok/sEstimated8B params
Run with ToolPiper

General · ibm-granite · 2026-04-06

Q8_0Excellent
10.3 GB11% of RAM~49 tok/sEstimated8.79B params
Run with ToolPiper

Multimodal · Google · 2026-03-02

Q8_0Excellent
6.2 GB6% of RAM~84 tok/sEstimated5.12B params
Run with ToolPiper

Multimodal · google · 2026-05

Q8_0Excellent
13.8 GB14% of RAM~36 tok/sEstimated11.96B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
1.5 GB2% of RAM~493 tok/sEstimated0.87B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
3.0 GB3% of RAM~189 tok/sEstimated2.27B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-24

Q8_0Excellent
40.6 GB42% of RAMBenchmark needed35.95B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
1.5 GB2% of RAM~493 tok/sEstimated0.87B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-28

Q8_0Excellent
3.0 GB3% of RAM~189 tok/sEstimated2.27B params
Run with ToolPiper

General · yuxinlu1 · 2026-06-28

Q8_0Excellent
13.8 GB14% of RAM~36 tok/sEstimated11.96B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-27

Q8_0Excellent
5.7 GB6% of RAM~92 tok/sEstimated4.66B params
Run with ToolPiper

General · Liquid AI · 2026-03-31

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

General · ravichandranj · 2026-02-13

Q8_0Excellent
7.7 GB8% of RAM~67 tok/sEstimated6.43B params
Run with ToolPiper

General · ibm-granite · 2026-04-16

Q8_0Excellent
9.8 GB10% of RAM~51 tok/sEstimated8.38B params
Run with ToolPiper

General · nvidia · 2026-03-18

Q8_0Excellent
35.7 GB37% of RAMBenchmark needed31.58B params
Run with ToolPiper

General · obliteratus · 2026-06-05

Q8_0Excellent
13.8 GB14% of RAM~36 tok/sEstimated11.96B params
Run with ToolPiper

General · sarvamai · 2026-03-03

Q8_0Excellent
36.4 GB38% of RAMBenchmark needed32.15B params
Run with ToolPiper

General · huihui-ai · 2026-07-11

Q8_0Excellent
13.8 GB14% of RAM~36 tok/sEstimated11.96B params
Run with ToolPiper

General · obliteratus · 2026-04-15

Q8_0Excellent
9.4 GB10% of RAM~54 tok/sEstimated8B params
Run with ToolPiper

General · nvidia · 2026-03-02

Q8_0Excellent
4.8 GB5% of RAM~112 tok/sEstimated3.83B params
Run with ToolPiper

General · Liquid AI · 2026-02-24

Q8_0Excellent
27.1 GB28% of RAMBenchmark needed23.84B params
Run with ToolPiper

Reasoning · jackrong · 2026-03-16

Q8_0Excellent
11.3 GB12% of RAM~44 tok/sEstimated9.65B params
Run with ToolPiper

Reasoning · jackrong · 2026-03-07

Q8_0Excellent
40.6 GB42% of RAMBenchmark needed35.95B params
Run with ToolPiper

General · cagrigungor · 2026-08-06

Q8_0Excellent
0.8 GB1% of RAM~1,589 tok/sEstimated0.27B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-27

Q8_0Excellent
11.3 GB12% of RAM~44 tok/sEstimated9.65B params
Run with ToolPiper

General · zai-org · 2026-01-19

Q8_0Excellent
35.3 GB37% of RAMBenchmark needed31.22B params
Run with ToolPiper

Multimodal · Alibaba · 2026-02-26

Q8_0Excellent
11.3 GB12% of RAM~44 tok/sEstimated9.65B params
Run with ToolPiper

General · nvidia · 2026-03-18

Q8_0Excellent
10.0 GB10% of RAM~51 tok/sEstimated8.49B params
Run with ToolPiper

Multimodal · Liquid AI · 2026-01-05

Q8_0Excellent
2.3 GB2% of RAM~268 tok/sEstimated1.6B params
Run with ToolPiper

General · Liquid AI · 2026-01-20

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

General · ai21labs · 2026-01-06

Q8_0Excellent
3.9 GB4% of RAMBenchmark needed3.03B params
Run with ToolPiper

Reasoning · openonerec · 2026-06-09

Q8_0Excellent
1.4 GB1% of RAM~536 tok/sEstimated0.8B params
Run with ToolPiper

General · Liquid AI · 2026-01-05

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

General · Liquid AI · 2026-01-04

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

General · Liquid AI · 2025-12-25

Q8_0Excellent
3.4 GB4% of RAM~167 tok/sEstimated2.57B params
Run with ToolPiper

General · Liquid AI · 2026-01-05

Q8_0Excellent
3.4 GB4% of RAM~167 tok/sEstimated2.57B params
Run with ToolPiper

Multimodal · davidau · 2026-02-02

Q8_0Excellent
5.3 GB6% of RAM~100 tok/sEstimated4.3B params
Run with ToolPiper

Chat · bineric · 2026-01-12

Q8_0Excellent
34.6 GB36% of RAMBenchmark needed30.53B params
Run with ToolPiper

General · inclusionai · 2026-02-09

Q8_0Excellent
18.6 GB19% of RAMBenchmark needed16.26B params
Run with ToolPiper

General · Liquid AI · 2025-10-28

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

General · allenai · 2025-11-18

Q8_0Excellent
8.6 GB9% of RAM~59 tok/sEstimated7.3B params
Run with ToolPiper

General · allenai · 2026-01-28

Q8_0Excellent
8.8 GB9% of RAM~58 tok/sEstimated7.43B params
Run with ToolPiper

Chat · pearl-ai · 2026-02-26

Q8_0Excellent
9.5 GB10% of RAM~53 tok/sEstimated8.03B params
Run with ToolPiper

Multimodal · Liquid AI · 2025-10-22

Q8_0Excellent
3.8 GB4% of RAM~143 tok/sEstimated3B params
Run with ToolPiper

General · frontiersmind · 2026-08-03

Q8_0Excellent
1.2 GB1% of RAM~660 tok/sEstimated0.65B params
Run with ToolPiper

General · alibaba-nlp · 2026-03-31

Q8_0Excellent
9.6 GB10% of RAM~52 tok/sEstimated8.19B params
Run with ToolPiper

General · openai · 2025-08-04

Q8_0Excellent
24.5 GB26% of RAMBenchmark needed21.51B params
Run with ToolPiper

Coding · Alibaba · 2025-07-31

Q8_0Excellent
34.6 GB36% of RAMBenchmark needed30.53B params
Run with ToolPiper

Chat · Liquid AI · 2026-01-06

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

General · typhoon-ai · 2025-09-23

Q8_0Excellent
5.0 GB5% of RAM~107 tok/sEstimated4.02B params
Run with ToolPiper

General · Alibaba · 2025-08-05

Q8_0Excellent
5.0 GB5% of RAM~107 tok/sEstimated4.02B params
Run with ToolPiper

General · ibm-granite · 2025-09-16

Q8_0Excellent
8.2 GB9% of RAMBenchmark needed6.94B params
Run with ToolPiper

General · ibm-granite · 2025-09-16

Q8_0Excellent
4.3 GB4% of RAM~126 tok/sEstimated3.4B params
Run with ToolPiper

Multimodal · Liquid AI · 2025-08-12

Q8_0Excellent
1.0 GB1% of RAM~953 tok/sEstimated0.45B params
Run with ToolPiper

General · allenai · 2025-09-12

Q8_0Excellent
8.6 GB9% of RAM~59 tok/sEstimated7.3B params
Run with ToolPiper

General · openai · 2025-09-18

Q8_0Excellent
24.5 GB26% of RAMBenchmark needed21.51B params
Run with ToolPiper

General · anton-hugging · 2026-02-06

Q8_0Excellent
9.0 GB9% of RAM~56 tok/sEstimated7.62B params
Run with ToolPiper

General · z-lab · 2026-01-04

Q8_0Excellent
9.4 GB10% of RAM~54 tok/sEstimated8B params
Run with ToolPiper

General · Alibaba · 2025-07-29

Q8_0Excellent
34.6 GB36% of RAMBenchmark needed30.53B params
Run with ToolPiper

General · ibm-granite · 2025-09-16

Q8_0Excellent
36.4 GB38% of RAMBenchmark needed32.21B params
Run with ToolPiper

General · openonerec · 2025-12-30

Q8_0Excellent
2.9 GB3% of RAM~201 tok/sEstimated2.13B params
Run with ToolPiper

General · ibm-granite · 2025-09-16

Q8_0Excellent
4.1 GB4% of RAM~134 tok/sEstimated3.19B params
Run with ToolPiper

General · bytedance · 2025-10-28

Q8_0Excellent
2.1 GB2% of RAM~300 tok/sEstimated1.43B params
Run with ToolPiper

General · dreamfast · 2026-01-11

Q8_0Excellent
14.1 GB15% of RAM~35 tok/sEstimated12.19B params
Run with ToolPiper

General · Liquid AI · 2025-10-07

Q8_0Excellent
9.8 GB10% of RAMBenchmark needed8.34B params
Run with ToolPiper

General · z-lab · 2026-01-04

Q8_0Excellent
5.0 GB5% of RAM~107 tok/sEstimated4B params
Run with ToolPiper

Multimodal · Liquid AI · 2025-08-12

Q8_0Excellent
2.3 GB2% of RAM~272 tok/sEstimated1.58B params
Run with ToolPiper

General · Liquid AI · 2025-09-22

Q8_0Excellent
3.4 GB4% of RAM~167 tok/sEstimated2.57B params
Run with ToolPiper

General · Liquid AI · 2025-09-30

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

General · Liquid AI · 2025-08-22

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

General · Liquid AI · 2025-09-03

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

General · Liquid AI · 2025-09-03

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

General · Liquid AI · 2025-08-25

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

General · Liquid AI · 2025-09-03

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

General · Liquid AI · 2025-09-03

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

General · NCAI · 2025-12-29

Q8_0Excellent
22.4 GB23% of RAMBenchmark needed19.6B params
Run with ToolPiper

General · NCAI · 2025-12-29

Q8_0Excellent
8.6 GB9% of RAMBenchmark needed7.25B params
Run with ToolPiper

Reasoning · DeepSeek · 2025-05-29

Q8_0Excellent
9.6 GB10% of RAM~52 tok/sEstimated8.19B params
Run with ToolPiper

Chat · allenai · 2025-11-19

Q8_0Excellent
8.6 GB9% of RAM~59 tok/sEstimated7.3B params
Run with ToolPiper

General · t-tech · 2025-12-22

Q8_0Excellent
9.6 GB10% of RAM~52 tok/sEstimated8.19B params
Run with ToolPiper

General · hmellor · 2025-07-22

Q8_0Excellent
1.9 GB2% of RAM~346 tok/sEstimated1.24B params
Run with ToolPiper

Chat · allenai · 2025-11-17

Q8_0Excellent
8.6 GB9% of RAM~59 tok/sEstimated7.3B params
Run with ToolPiper

General · nvidia · 2025-08-12

Q8_0Excellent
10.4 GB11% of RAM~48 tok/sEstimated8.89B params
Run with ToolPiper

General · inclusionai · 2025-11-25

Q8_0Excellent
18.6 GB19% of RAMBenchmark needed16.26B params
Run with ToolPiper

General · ibm-granite · 2025-04-30

Q8_0Excellent
7.9 GB8% of RAMBenchmark needed6.67B params
Run with ToolPiper

General · Liquid AI · 2025-07-10

Q8_0Excellent
1.8 GB2% of RAM~367 tok/sEstimated1.17B params
Run with ToolPiper

Reasoning · Microsoft · 2025-04-29

Q8_0Excellent
4.8 GB5% of RAM~112 tok/sEstimated3.84B params
Run with ToolPiper

Chat · xcuros · 2026-02-28

Q8_0Excellent
9.0 GB9% of RAM~56 tok/sEstimated7.62B params
Run with ToolPiper

General · amd · 2025-05-17

Q8_0Excellent
2.2 GB2% of RAM~286 tok/sEstimated1.5B params
Run with ToolPiper

General · stefanruseti · 2025-06-04

Q8_0Excellent
1.9 GB2% of RAM~346 tok/sEstimated1.24B params
Run with ToolPiper

General · lgai-exaone · 2025-07-11

Q8_0Excellent
1.9 GB2% of RAM~335 tok/sEstimated1.28B params
Run with ToolPiper

General · Liquid AI · 2025-07-10

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

General · menlo · 2025-06-25

Q8_0Excellent
5.0 GB5% of RAM~107 tok/sEstimated4.02B params
Run with ToolPiper

General · Liquid AI · 2025-07-10

Q8_0Excellent
1.3 GB1% of RAM~580 tok/sEstimated0.74B params
Run with ToolPiper

Multimodal · NCAI · 2025-12-29

Q8_0Excellent
9.0 GB9% of RAMBenchmark needed7.58B params
Run with ToolPiper

Multimodal · Alibaba · 2025-01-26

Q8_0Excellent
4.7 GB5% of RAM~114 tok/sEstimated3.75B params
Run with ToolPiper

General · huggingfacetb · 2025-07-08

Q8_0Excellent
3.9 GB4% of RAM~139 tok/sEstimated3.08B params
Run with ToolPiper

General · Alibaba · 2025-09-23

Q8_0Excellent
5.4 GB6% of RAM~97 tok/sEstimated4.41B params
Run with ToolPiper

General · nvidia · 2025-08-21

Q8_0Excellent
13.9 GB14% of RAM~36 tok/sEstimated12B params
Run with ToolPiper

General · k-chirkunov · 2026-07-14

Q8_0Excellent
9.4 GB10% of RAM~54 tok/sEstimated7.94B params
Run with ToolPiper

General · avitotech · 2025-10-20

Q8_0Excellent
9.3 GB10% of RAM~54 tok/sEstimated7.9B params
Run with ToolPiper

General · huggingfacetb · 2025-06-19

Q8_0Excellent
3.9 GB4% of RAM~139 tok/sEstimated3.08B params
Run with ToolPiper

General · Alibaba · 2025-09-23

Q8_0Excellent
1.3 GB1% of RAM~572 tok/sEstimated0.75B params
Run with ToolPiper

General · nvidia · 2025-03-16

Q8_0Excellent
9.5 GB10% of RAM~53 tok/sEstimated8.03B params
Run with ToolPiper

General · dream-org · 2025-04-03

Q8_0Excellent
9.0 GB9% of RAM~56 tok/sEstimated7.62B params
Run with ToolPiper

General · openbmb · 2025-09-02

Q8_0Excellent
9.6 GB10% of RAM~52 tok/sEstimated8.19B params
Run with ToolPiper

General · Alibaba · 2025-09-23

Q8_0Excellent
9.6 GB10% of RAM~52 tok/sEstimated8.19B params
Run with ToolPiper

General · pfnet · 2025-02-05

Q8_0Excellent
1.9 GB2% of RAM~333 tok/sEstimated1.29B params
Run with ToolPiper

General · bytedance-seed · 2025-04-09

Q8_0Excellent
11.0 GB11% of RAMBenchmark needed9.37B params
Run with ToolPiper

Chat · baseten · 2025-09-12

Q8_0Excellent
4.1 GB4% of RAM~134 tok/sEstimated3.21B params
Run with ToolPiper

General · sapientinc · 2026-05-17

Q8_0Excellent
1.8 GB2% of RAM~364 tok/sEstimated1.18B params
Run with ToolPiper

General · ibm-granite · 2025-10-07

Q8_0Excellent
0.9 GB1% of RAM~1,226 tok/sEstimated0.35B params
Run with ToolPiper

Chat · utter-project · 2026-01-26

Q8_0Excellent
10.7 GB11% of RAM~47 tok/sEstimated9.15B params
Run with ToolPiper

General · fableforge-ai · 2026-07-05

Q8_0Excellent
2.2 GB2% of RAM~279 tok/sEstimated1.54B params
Run with ToolPiper

General · viorikaai-org · 2026-07-05

Q8_0Excellent
0.6 GB1% of RAM~8,580 tok/sEstimated0.05B params
Run with ToolPiper

General · bananamind · 2026-07-17

Q8_0Excellent
0.5 GB1% of RAM~42,900 tok/sEstimated0.01B params
Run with ToolPiper

60-core vs 80-core GPU

Both bins run the same 819 GB/s memory bus, so token generation is the same on either one. The extra cores show up in image and video work, not in tokens per second.

ConfigurationMemory bandwidthMemory optionsModels that fit
28-core CPU, 60-core GPU819 GB/s96 GBIdentical
32-core CPU, 80-core GPU819 GB/s96, 256, 512 GBIdentical

Measured on the M3 Ultra

Nobody has submitted a benchmark on the M3 Ultra yet, so every speed on this page is the formula estimate rather than a measured run. The estimate is bandwidth-driven and calibrated against chips that do have data, which makes it a good guide and not a promise.

ToolPiper contributes a result anonymously when you run the benchmark, and the leaderboard shows every chip that already has one.

What comes next

Projection dated 2026-09-08

Apple has not announced any of this. The M7 Max and Ultra are the first parts Apple designed after cancelling a generation to reach them, so extrapolating from the M5 under-represents them. These rows assume LPDDR6, whose wider channels grow every bus by half, at its top speed bin by the time the Max and Ultra ship. The 1.5 TB Ultra ceiling is Bloomberg's reported design target, and whether that configuration ships depends on the memory market. Stacked memory or a new package fabric would land above these numbers; nobody outside Apple can price that yet.

M7 UltraProjected

Projected chip · expected 2028

2765 GB/s · 256 to 1536 GB unified memory

40-core CPU · 112 or 128-core GPU · 152 TOPS Neural Engine

Would hold about a 2092B model at Q4

Why unified memory is the number that matters

On a PC the model has to fit in GPU VRAM, which is a separate pool from system RAM and usually the smaller of the two. Apple Silicon has one pool. The M3 Ultra's 819 GB/s bus is shared by CPU, GPU, and Neural Engine, so a 512 GB machine can hand almost all of that to a model with no copy across a bus.

Apple stopped selling this one, which is exactly why it is interesting. The Studio exists for this workload. It carries the widest memory buses and the highest capacities Apple sells, and it runs at full clocks indefinitely. A used M3 Ultra at 512 GB still gives you 819 GB/s and a hard 696B ceiling, and neither number degrades with age the way a battery does.

Common questions

Can the M3 Ultra Mac Studio run a 70B model?

Yes, at 512 GB. A 70B model at Q4_K_M needs about 46 GB including an 8K context, and 512 GB of unified memory leaves about 450 GB for weights once macOS takes its share. At 96 GB it does not fit at any quantization worth running.

How much unified memory should I get with the M3 Ultra Mac Studio?

Memory is the only spec that changes what you can run at all. 96 GB holds about a 129B model at Q4; 512 GB holds about 696B. It is soldered, so this is a one-time decision, and it is the upgrade worth paying for before core count.

How fast are local LLMs on the M3 Ultra?

Token generation is bandwidth-bound, so M3 Ultra throughput scales with its 819 GB/s memory bus. Divide bandwidth by the size of the weights actually read per token to get the ceiling, then expect roughly half of that in practice. A 7B model at Q4 reads about 4 GB per token pass, so the M3 Ultra lands in the tens of tokens per second and a 70B model lands in the single digits.

Is the 80-core GPU worth it over the 60-core on the M3 Ultra?

Not for LLMs. Both bins run the same 819 GB/s memory bus and take the same memory options, and token generation is bound by bandwidth rather than GPU cores. The extra cores show up in image generation and video work, not in tokens per second.

Is a used M3 Ultra Mac Studio still worth buying for local AI?

For inference, the specs that matter do not age: 819 GB/s and up to 512 GB of unified memory are the same numbers today as they were in 2025. A used M3 Ultra at the top memory option usually beats a new base-tier machine at the same price on both. Check the battery and the display, not the silicon.

Run these models on your Mac Studio

ToolPiper downloads, manages, and runs local models on Apple Silicon. Free, and nothing leaves the machine.