Mac Studio M5 Ultra · 512 GB
512 GB of unified memory at 1.2 TB/s — 400 GB-class frontier weights in one near-silent box, back on sale since October.
Price
£22,299 inc VAT
Apple UK — 512 GB only with the 36C/80C M5 Ultra (£6,799 + £15,500) · orderable since 28 Oct 2026
verified 2026-11 · supply & lease options in every proposal
Apple- 512 GB
- Unified memory
- 1.2 TB/s
- Memory bandwidth
- 80-core
- GPU
- 385 W
- Max power
one coherent pool
frontier weights at usable speed
36-core CPU — required for 512 GB
Apple-measured on this configuration
The configuration that made the Mac Studio the folk hero of local AI is back. The M3 Ultra's 512 GB option vanished in the March 2026 memory shortage; the M5 Ultra restores it (orderable since 28 October 2026), paired only with the top 36-core CPU / 80-core GPU chip, at £22,299 inc VAT with a 1 TB SSD.
512 GB at 1.2 TB/s holds 400 GB-class open weights whole — GLM-5.3 at 4-bit, with little room to spare, or MiniMax M3.1 Flash with plenty — and Apple's 385 W maximum was measured on exactly this configuration. Prompt processing runs ~3.6× the M3 Ultra's.
It is no longer the value anomaly it was at $9,499: £22,299 buys more than five times the memory of a single RTX PRO 6000 workstation, but no CUDA. Used M3 Ultra 512 GB units now ask $19–23k on eBay, so new is the sensible buy — and Thunderbolt 5 RDMA pairs scale it to trillion-parameter class.
vs its siblings: The only near-silent desk box that holds 400 GB-class frontier weights whole.
Memory, to scale
512 GB model-visible · bandwidth is the speed limit
Unified memory
512 GB · 1.2 TB/s
LPDDR5x
For scale
DGX Spark — 128 GB @ 273 GB/s
RTX PRO 6000 — 96 GB @ 1.79 TB/s
DGX Station GB300 — 748 GB coherent
Thunderbolt 5 RDMA — two 512 GB Studios pool ~1 TB for trillion-parameter 4-bit weights; Apple quotes up to 3× faster inference from four.
What it actually runs
Declared from research and benchmarks, not computed marketing — tokens-per-second figures are cited where a real measurement exists.
- GLM-5.34-bit (~464 GB)fitsjoint-strongest downloadable model, on a desk — little context room
- MiniMax M3.1 FlashQAT 4-bit (252 GB)with headroomfrontier-class multimodal with real context headroom
- DeepSeek V4 Flashnative FP4/FP8 (167 GB)with headroomplenty of room for long context
- Qwen3.8-Flash-NextFP8 (186 GB)with headroomnear-full precision with room to spare
- Mistral Large 44-bit (610 GB)2+ units linkedtwo Studios over TB5 RDMA
- gpt-oss-120bMXFP4with headroom~115–120 tok/s — with 440 GB to spare
The full sheet
Compute
- Chip
- Apple M5 Ultra — 36-core CPU, 80-core GPU (required for 512 GB)
- Neural Engine
- 32-core · Neural Accelerators in every GPU core
Memory
- Unified memory
- 512 GB @ 1.2 TB/s (no 768 GB option)
- Model capacity
- 400 GB-class weights at 4-bit — GLM-5.3 (~464 GB) fits, tightly
Storage & I/O
- SSD
- 1 TB base, to 16 TB
- Ports
- 6× Thunderbolt 5 · 10 GbE
Status
- Availability
- Orderable since 28 Oct 2026 (Apple UK)
- Previous generation
- M3 Ultra 512 GB discontinued Mar 2026 — used units ~$19–23k asking
Where it shines
- 400 GB-class frontier weights in one near-silent box
- 1.2 TB/s bandwidth and ~3.6× the M3 Ultra's prompt processing
- Under 400 W at full load — no facilities conversation required
- TB5 RDMA pairs reach trillion-parameter class
The trade-offs
- £22,299 — 512 GB forces the top chip and a £15,500 memory upgrade
- No CUDA ecosystem; MLX/llama.cpp serving only
- No 768 GB option — trillion-parameter models still need two
Buy this box for
Understanding Apple Mac Studio
Apple M5 Ultra
Apple silicon's unified-memory architecture gives the GPU the entire RAM pool. The M5 Ultra Mac Studio, shipping since 22 September 2026, runs that pool at 1.2 TB/s — about 4.4× a DGX Spark — with Neural Accelerators in every GPU core. Measured on a 256 GB unit, gpt-oss-120b generates at ~115–120 tok/s and dense 70B-class models at ~23 tok/s, and Apple measures the top configuration at 385 W flat out.
The big configurations are back after the 2026 memory shortage withdrew the M3 Ultra's 256 GB and 512 GB options: the M5 Ultra starts at £5,499 with 96 GB, adds 256 GB for £4,000, and offers 512 GB (with the top 36-core CPU / 80-core GPU chip) at £22,299 — orderable since 28 October. There is no 768 GB option. Clustering still works: macOS RDMA over Thunderbolt 5 pools several Studios, and Apple quotes up to 3× faster inference from a four-machine cluster.
Prompt processing, long the platform's weakness, has improved sharply — about 3.6× the M3 Ultra on a 24k-token prompt (Apple claims up to 4×). There is still no CUDA, so you serve through MLX, llama.cpp, LM Studio or Ollama rather than vLLM: the right machine for chat, document and agent workloads, less so for high-concurrency serving.
Sources & verification
Specifications and prices verified 2026-11 against the sources below. The memory shortage is repricing this market monthly — we re-verify at quote.
Compare against the rest of the catalogue, or have us spec it against your workloads.
Back to catalogueSpec the Mac Studio M5 Ultra · 512 GB for your workload
We'll size it against your real token volumes, quote supply and lease options, and stand up the models with evals that prove they do the job — or tell you honestly that a different box (or the API) wins.