ailocalinference

← All hardware

Mac Studio M5 Ultra

Complete box · $5,499/96GB · $9,499/256GB · 512GB late Oct, price TBD (~$13.5k est)

AppleComplete system$13,500 (estimated) · Aug 31, 2026View on Amazonpaid link

Memory
512 GB~384 GB usable
Bandwidth
1.2kGB/s — sets decode speed
Compute
60FP16 TFLOPS — sets prefill
Power
270 Wunder sustained load
Noise
35 dBAdesk-friendly
Price per GB
$26of accelerator memory

Price history

tracking

Tracking since 2026-08-31 — not enough history for a trend yet.

1 observation recorded · Last checked 2026-08-31. Check the retailer before acting on this — we would rather say the data is old than pretend otherwise.

What it can run

5 of 6 reference workloads fit. The largest is 320B MoE A18B · FP8 (frontier-ish), at an estimated 28 tok/s of single-stream decode — the fastest of the 2 machines here that can run it.

WorkloadNeedsFitsPrefill t/sDecode t/sBatched t/s$ per t/s
8B dense · Q4 (chat, drafts)
dense · 4.7 GB read per token
5.4 GBYes563166404$81
32B dense · Q4 (coding)
dense · 19 GB read per token
21.8 GBYes14141101$329
70B dense · Q4 (strong general)
dense · 40 GB read per token
46 GBYes642046$675
120B MoE A5B · MXFP4 (gpt-oss class)
MoE · 2.9 GB read per token
72.4 GBYes882174599$78
320B MoE A18B · FP8 (frontier-ish)
MoE · 18 GB read per token
368 GBYes25028146$482
671B MoE A37B · Q4 (DeepSeek class)
MoE · 18.5 GB read per token
437 GBNo1222784—

Figures are modelled from this machine's bandwidth and compute, not measured on one. How they are derived →

Alternatives that also run 320B MoE A18B · FP8 (frontier-ish)

MachinePriceMemoryDecode t/s$ per t/s
Mac Studio M5 Ultra — this machine
$13,500512 GB28$482
$9,499512 GB19$500